Cranial ontogenetic variation in early saurischians and the role of heterochrony in the diversification of predatory dinosaurs Christian Foth1,2,3, Brandon P. Hedrick4,5 and Martin D. Ezcurra2,6,7 1 SNSB, Bayerische Staatssammlung für Paläontologie und Geologie, München, Germany 2 Department of Earth and Environmental Sciences, Ludwig-Maximilians-Universität, München, Germany 3 Department of Geosciences, University of Fribourg/Freiburg, Fribourg, Switzerland 4 Department of Earth and Environmental Science, University of Pennsylvania, Philadelphia, PA, United States 5 Department of Biology, University of Massachusetts, Amherst, MA, United States 6 CONICET, Sección Paleontología de Vertebrados, Museo Argentino de Ciencias Naturales, Buenos Aires, Argentina 7 School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham, United Kingdom

ABSTRACT Non-avian saurischian skulls underwent at least 165 million years of evolution and shapes varied from elongated skulls, such as in the theropod Coelophysis, to short and box-shaped skulls, such as in the sauropod Camarasaurus. A number of factors have long been considered to drive skull shape, including phylogeny, dietary preferences and functional constraints. However, heterochrony is increasingly being recognized as an important factor in dinosaur evolution. In order to quantitatively analyse the impact of heterochrony on saurischian skull shape, we analysed five ontogenetic trajectories using two-dimensional geometric morphometrics in a phylogenetic framework. This allowed for the comparative investigation of main ontogenetic shape changes and the evaluation of how heterochrony affected skull shape through both ontogenetic and phylogenetic trajectories. Using principal component analyses and multivariate regressions, it was possible to quantify different ontogenetic trajectories and evaluate Submitted 17 June 2015 them for evidence of heterochronic events allowing testing of previous hypotheses on Accepted 21 December 2015 Published 18 January 2016 cranial heterochrony in saurischians. We found that the skull shape of the hypothetical ancestor of Saurischia likely led to basal Sauropodomorpha through paedomorphosis, Corresponding author Christian Foth, chris- and to basal Theropoda mainly through peramorphosis. Paedomorphosis then led [email protected] from Orionides to Avetheropoda, indicating that the paedomorphic trend found by Academic editor previous authors in advanced coelurosaurs may extend back into the early evolution of Jesús Marugán-Lobón Avetheropoda. Not only are changes in saurischian skull shape complex due to the large Additional Information and number of factors that affected it, but heterochrony itself is complex, with a number of Declarations can be found on possible reversals throughout non-avian saurischian evolution. In general, the sampling page 28 of complete ontogenetic trajectories including early juveniles is considerably lower than DOI 10.7717/peerj.1589 the sampling of single adult or subadult individuals, which is a major impediment to the study of heterochrony on non-avian dinosaurs. Thus, the current work represents Copyright 2016 Foth et al. an exploratory analysis. To better understand the cranial ontogeny and the impact of heterochrony on skull evolution in saurischians, the data set that we present here must Distributed under be expanded and complemented with further sampling from future fossil discoveries, Creative Commons CC-BY 4.0 especially of juvenile individuals. OPEN ACCESS

How to cite this article Foth et al. (2016), Cranial ontogenetic variation in early saurischians and the role of heterochrony in the diversifi- cation of predatory dinosaurs. PeerJ 4:e1589; DOI 10.7717/peerj.1589 Subjects Biodiversity, Evolutionary Studies, Paleontology, Taxonomy, Zoology Keywords Ontogeny, Sauropodomorpha, Evolution, Theropoda, Skull shape, Dinosauria, Heterochrony, Geometric morphometrics INTRODUCTION In an evolutionary context, heterochrony describes phenotypic changes due to shifts in the timing or rate of developmental processes in an organism relative to its ancestor, and can lead to significant evolutionary changes in body plans within relatively short periods of time (Gould, 1977; Alberch et al., 1979; McNamara, 1982; Reilly, Wiley & Meinhardt, 1997; Klingenberg, 1998; McNamara & McKinney, 2005). Two major types of heterochronic processes are discerned: paedomorphosis and peramorphosis. Paedo- morphosis occurs when the later ontogenetic stages of an organism retain characteristics from earlier ontogenetic stages of its ancestor due to a truncation of the growth period (progenesis), decrease of the growth rate (neoteny) or a delayed onset of developmental processes (postdisplacement). In contrast, a peramorphic organism is ontogenetically more developed than the later ontogenetic stages of its ancestor due to the extension of growth period (hypermorphosis), the increase of the growth rate (acceleration) or the earlier onset of developmental processes (predisplacement) (see Gould, 1977; Alberch et al., 1979; Klingenberg, 1998). In practice, evidence for heterochronic events in evolution can be detected by comparing the ontogenetic trajectories of different taxa under the consideration of their phylogenetic interrelationships (Alberch et al., 1979; Fink, 1982). Thus, the concept of heterochrony connects two main fields of biological sciences: developmental and evolutionary biology (Gould, 1977; Raff, 1996). When studying heterochrony, ontogenetic trajectories are characterized by three separate vectors (size, shape, and ontogenetic age), which allows for quantification of heterochronic processes with slope, length and position within a Euclidean space (Alberch et al., 1979). In this context, geometric morphometrics is a useful method for characterizing shape and size vectors to investigate heterochrony in organisms within a multivariate framework (Mitteroecker, Gunz & Bookstein, 2005). Documentation of heterochrony in the vertebrate fossil record is limited. Preserved fossil ontogenetic series covering the whole postnatal development of fossil species are rare due to the fact that early juvenile specimens are often either lacking or incomplete. Furthermore, exact ages of single ontogenetic stages are often not available, resulting in the temporal component often being replaced by size, which is not an ideal variable for age (Klingenberg, 1998; Gould, 2000). Nevertheless, the role of heterochrony has been recognized and discussed for the evolution of multiple fossil lineages that do preserve ontogenetic series (Balanoff & Rowe, 2007; Gerber, Neige & Eble, 2007; Schoch, 2009; Schoch, 2010; Schoch, 2014; Bhullar, 2012; Forasiepi & Sánchez-Villagra, 2014; Ezcurra & Butler, 2015), including non-avian dinosaurs (e.g., Long & McNamara, 1997; Erickson et al., 2004; Guenther, 2009; Bhullar et al., 2012; Canale et al., 2014). For example, Long & McNamara (1997), Erickson et al. (2004) and Canale et al. (2014) hypothesized that the evolution of large body size in carcharodontosaurids and tyrannosaurids from medium-sized ancestors was the result of peramorphosis.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 2/41 There has recently been an increasing interest in shape diversity in non-avian dinosaurs, in which geometric morphometric methods have been applied on a regular basis (e.g., Bonnan, 2004; Chinnery, 2004; Campione & Evans, 2011; Hedrick & Dodson, 2013; Lautenschlager, 2014; Schwarz-Wings & Böhm, 2014; Maiorini et al., 2015). Skull shape diversity in saurischian dinosaurs has been studied in particular detail (e.g., Henderson, 2002; Young & Larvan, 2010; Rauhut et al., 2011; Brusatte et al., 2012; Bhullar et al., 2012; Foth & Rauhut, 2013a; Foth & Rauhut, 2013b), but usually in relation to functional constraints, dietary preferences, phylogenetic interrelationships, and macroevolutionary patterns. For example, these studies have shown that skull shape in sauropodomorphs and theropods is phylogenetically constrained (Young & Larvan, 2010; Brusatte et al., 2012; Foth & Rauhut, 2013a) and that the shape of the orbit in theropods is functionally constrained (Henderson, 2002; Foth & Rauhut, 2013a). Thus, geometric morphometrics is a powerful method to quantify both intraspecific (e.g., ontogeny, sexual dimorphism, polymorphism) and interspecific (e.g., systematics, macroevolution) shape variation on the basis of homologous landmarks or outlines, which capture more information about shape than traditional morphometric measurements (Corti, 1993; Rohlf & Marcus, 1993; Adams, Rohlf & Slice, 2004; Adams, Rohlf & Slice, 2013; Slice, 2007; Mitteroecker & Gunz, 2009; Zelditch, Swiderski & Sheets, 2012). As a result, geometric morphometrics has also been successfully applied to the study of heterochrony among various tetrapod groups, in which the univariate mathematical approach of Alberch et al. (1979) was adapted to a multivariate framework (e.g., Berge & Penin, 2004; Mitteroecker et al., 2004; Mitteroecker, Gunz & Bookstein, 2005; Lieberman et al., 2007; Drake, 2011; Piras et al., 2011; Bhullar et al., 2012). However, only Bhullar et al. (2012) have examined cranial shape diversity of theropod dinosaurs using multivariate methods in the context of heterochrony. This pioneering study demonstrated that recent birds have highly paedomorphic skulls compared to non-avian theropods and Mesozoic birds (e.g., Archaeopteryx and Enantiornithes), which evolved in a multistep transformation within the clade Eumaniraptora. Furthermore, Bhullar et al. (2012) found evidence for independent peramorphic trends in the skull shape of large-bodied tyrannosaurids, dromaeosaurids and troodontids and proposed a similar trend for allosaurids. Finally, Bhullar et al. (2012) hypothesized a possible paedomorphosis for Eoraptor and basal sauropodomorphs. The aim of the current study is to investigate the cranial shape diversity of saurischian dinosaurs by comparing the ontogenetic trajectories of different taxa from both qualitative and quantitative data, using two-dimensional geometric morphometrics (2D GM). This study expands on the work of Bhullar et al. (2012) who focused primarily on trends within Maniraptora, derived non-avian theropods and basal avian theropods. We have built upon their study by including an improved sample of basal saurischians and theropods (including a number of different ontogenetic series), which should be more sensitive for testing of the heterochronic changes for allosaurids and basal sauropodomorphs proposed, but not verified statistically, by Bhullar et al. (2012). The phylogenetic relationships of the ontogenetic series sampled in this study are integrated into an ancestor-descendant framework to look for further potential heterochronic processes in the cranial evolution of saurischians. However, due to the limited number of ontogenetic series known for

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 3/41 sauropodomorphs, the current study focuses primarily on the early evolution of theropods. Nevertheless, due to the limited number of ontogenetic series currently available in our taxonomic sample, this work must be viewed as an exploratory study, which will need to be expanded and complemented with further sampling from future fossil discoveries.

MATERIALS AND METHODS Taxon sampling We sampled the crania of 35 saurischian dinosaur taxa (10 sauropodomorphs and 25 non- pennaraptoran theropods, see Table S3) on the basis of published reconstructions of adult (or advanced subadult) individuals in lateral view (with exception of the reconstructions of the basal tyrannosauroid Dilong [IVPP V14243] and the basal alvarezsauroid Haplocheirus [IVPP V15988], which were based on our personal observations). The data set shows an overlap of 15 terminal taxa with that of Bhullar et al. (2012) and builds on that study with an addition of 20 new taxa. Theropods with large nasal crests (e.g., Ceratosaurus, Dilophosaurus, Guanlong) were excluded from the primary data set as they were found to have a strong impact on the ancestral shape reconstruction (see below) of Averostra, Avetheropoda, Coelurosauria and Tyrannosauroidea (see Fig. S5 and Table S6). Although cranial crests are a common structure among theropod dinosaurs (Molnar, 2005), reconstruction of moderately to strongly crested hypothetical ancestors within this study would necessarily be artificial due to the lack of intermediate crested forms and relatively small sample size of the available data set. Only Monolophosaurus was included in the main data set because it possesses a rather moderately sized and simple nasal crest. ‘Syntarsus’ kayentakatae, which is often reconstructed with a pair of prominent nasal crests (Rowe, 1989; Tykoski, 1998), was analysed in this study without crests since this structure is probably artificial due to post-mortem displacement of the nasals (Ezcurra & Novas, 2005; Ezcurra & Novas, 2007). As cranial crests usually represent external visual signal structures (Sampson, 1999; Padian & Horner, 2011; Hone, Naish & Cuthill, 2012), their evolutionary development most likely represents either an evolutionary novelty or was sourced from regional peramorphic processes if the primordia were already present in the ancestor (see discussion on the evolution of horns and frills in Ceratopsia by Long & McNamara (1997)). However, we generated a second data set that includes crested taxa for comparison with the main data set (see below). In our sample, five taxa preserve early ontogenetic stages allowing the capture of both juvenile and adult skull shapes, which were used to reconstruct five simplified ontogenetic series, containing two stages (i.e., an early juvenile and adult stage). This sample includes the basal sauropodomorph Massospondylus, the basal theropod Coelophysis, the megalosaurid Dubreuillosaurus, the allosauroid Allosaurus, and the tyrannosaurid Tarbosaurus (see Table S4). Two of the ontogenetic series sampled (Coelophysis and Tyrannosaurus/Tar- bosaurus) overlap with the data set from Bhullar et al. (2012), but we expand on the previous study by including three more basal trajectories in order to concentrate on a different part of the theropod tree. As the fossil record of juvenile dinosaur specimens with complete skull material is rare, the number of ontogenetic series is limited. To improve

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 4/41 sampling, previous studies have included reconstructions from multiple partial juvenile skulls or juveniles from closely related taxa (e.g., Bhullar et al., 2012). We implemented this approach in two cases: the reconstruction of the juvenile Coelophysis sample was based on three incomplete, somewhat taphonomically deformed individuals (MCZ 4326; GR 392; CM 31375); and the holotype of Sciurumimus (BMMS BK 11) was used as the juvenile representative of the megalosaurid Dubreuillosaurus based on the phylogenetic analyses of Rauhut et al. (2012). In contrast to Bhullar et al. (2012), we did not include the ontogenetic series of Byronosaurus, Therizinosauridae (represented by a therizinosaurid embryo and the skull of Erlikosaurus) and Compsognathus (with the juvenile specimen represented by Scipionyx) in the data set because the postorbital region of the juvenile skulls of the former two taxa is crushed or incomplete (Bever & Norell, 2009; Kundrát et al., 2008), and the taxonomic referral of Scipionyx to the clade Compsognathidae (see Dal Sasso & Maganuco, 2011) is uncertain and maybe an artefact of coding juvenile character states (see Rauhut et al., 2012). Two-dimensional geometric morphometrics (2D GM) We used 20 landmarks (LMs) and 51 semi-landmarks (semi-LMs) on our sample in order to accurately capture skull shape. The landmarks were collected using the software tpsDig2 (Rohlf, 2005) and were classified as either type 1 (points where two bone sutures meet) or type 2 (points of maximum curvature and extremities) (Bookstein, 1991) (see Fig. S1 and Table S1 for full description). Type 3 landmarks (points constructed between two homologous landmarks, which mainly define the shape of the skull or skull openings rather than the position of exact homologous points) were not used in our study. Semi-landmarks were used to capture the shape of skull openings and the skull outline by defining a number of points that are placed equidistantly along respective curves (Bookstein, 1991; Bookstein et al., 1999). The percent error for digitizing each landmark and semi-landmark was estimated for the skull reconstruction of the juvenile Coelophysis (with n = 10 replications) using the method described by Singleton (2002). Landmark and semi-landmark error varies between 0.117 percent (LM 51—most posterior point of the descending process of the maxilla contacting the nasal and/or the lacrimal) and 0.738% (LM 3—contact between the maxilla and jugal along the ventral margin of the skull) with a mean of 0.283%. The error has no significant effect on the shape analyses (see Table S2). The shape coordinates were then imported into the software package MorphoJ 1.05d (Klingenberg, 2011) and superimposed using generalized Procrustes analysis (GPA). GPA rotates, translates and resizes landmark coordinates of all specimens accounting for all non-shape related differences between landmark configurations, leaving only shape information (Gower, 1975; Rohlf & Slice, 1990). Although semi-landmarks have fewer degrees of freedom than regular landmarks (and thus contain less shape information) (Bookstein, 1991), we treated landmarks and semi-landmarks as equivalent for GPA (Zelditch, Swiderski & Sheets, 2012) and did not slide the semi-landmarks. The sliding process created considerable artificial deformation on the Procrustes-fitted shape in some taxa (see Fig. S2). However, due to the equivalent weighting of landmarks and semi-landmarks, it should be kept in mind that the shape information captured by the

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 5/41 semi-landmarks strongly influences the results (Zelditch, Swiderski & Sheets, 2012; see below). In order to estimate the influence of the semi-landmarks on the shape data, all analyses described below were also applied to an additional data set that included only landmark data (see Supplemental Information). The generated Procrustes coordinates were used to compare juvenile and adult skull shapes to each other in each ontogenetic series to find ontogenetic patterns between and within taxa. Furthermore, the Procrustes coordinates of all taxa (including juvenile specimens) were subjected to an exploratory principal components analysis (PCA) using the covariance matrix generated from Procrustes coordinates. PCA simplifies descriptions of variation among individuals by creating new sets of variables that are linear combinations of the original set such that the new sets are independent from one another and have zero covariance. The principal components (PCs) describe successively smaller amounts of total variance of the sample. This allows for a larger proportion of the variance to be described using a smaller number of variables than the original data would have allowed (Zelditch, Swiderski & Sheets, 2012). A multivariate regression of the Procrustes coordinates against log-transformed centroid sizes (=square root of the sum of the squared distances of each landmark to the centroid of the landmark configuration, Zelditch, Swiderski & Sheets, 2012) was performed to test if the skull shape variation is correlated with size and contains allometric information (Drake & Klingenberg, 2008). Quantification of ontogenetic trajectories The different ontogenetic trajectories generated in the PCA and regression analyses were compared to each other by calculating pairwise two-dimensional angles between different trajectories based on the PC values of the first three axes, which are the significant principal components (significance calculated using the broken stick method, see Jackson, 1993). Each of the two-stage ontogenetic trajectories was described as a phenotypic change vector, 1yEi =y Eij −y Eik, with two shape traits (PC 1 vs. PC 2 and PC 1 vs. PC 3), where i is a specific ontogeny between two fixed stages, juvenile (j) and adult (k)(Collyer & Adams, 2007). The difference in direction (angle) between the ontogenetic phenotypic change vectors E · E 1yE ,1yE was calculated using the dot product cos−1(1yE ,1yE ) = 1ya 1yb . PC values a b a b |1yEa||1yEb| were employed to calculate the length of each ontogenetic trajectory. Lengths and angles were used to characterise the differences between the ontogenetic trajectories in relation to shape variation. Phylogenetic framework for heterochronic analyses In an evolutionary context, heterochrony is defined as the change in the timing or rate of developmental processes in ancestor-descendant relationships (Alberch et al., 1979; Fink, 1982; Klingenberg, 1998), and thus a direct comparison of ontogenetic trajectories from different species (as terminal taxa) can be problematic because it is hard to determine which trajectory would represent the ancestral and the descendant form, respectively (see Fink, 1982). This is exacerbated when the supposed ancestral (terminal) species possesses an unknown, long evolutionary history resulting from a ghost lineage. This problem can be partially solved using a phylogenetic approach (see Alberch et al., 1979;

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 6/41 Fink, 1982; Balanoff & Rowe, 2007; Bhullar, 2012; Fritsch, Bininda-Emonds & Richter, 2013; Mallon, Ryan & Campbell, 2015), in which the ancestor of two sister (terminal) taxa is represented by the hypothetical last common ancestor (Hennig, 1966). Therefore, on the basis of the phylogenetic distribution of the five ontogenetic series sampled we calculated hypothetical ancestral ontogenetic trajectories for Saurischia, Neotheropoda, Orionides and Avetheropoda using ancestral shape reconstructions as follows (see Figs. S3 and S4). An informal supertree (sensu Butler & Goswami, 2008) including all taxa with adult individuals was created based on recent phylogenetic analyses (see Figs. S3 and S4): basal Sauropodomorpha (Cabreira et al., 2011), Coelophysoidea (Ezcurra & Novas, 2007), Ceratosauria (Pol & Rauhut, 2012), Tetanurae (Carrano, Benson & Sampson, 2012), and Coelurosauria (Turner, Makovicky & Norell, 2012; Loewen et al., 2013). The phylogenetic position of Eoraptor follows Martínez et al. (2011) and Martínez, Apaldetti & Abelin (2013). The position of Adeopapposaurus as sister taxon of Massospondylus follows Martínez (2009). The position of Herrerasaurus and Tawa at the base of Theropoda is based on Sues et al. (2011). Zupaysaurus was placed outside Coelophysoidea as one of the successive sister taxa of Averostra (Smith et al., 2007; Sues et al., 2011; Ezcurra, 2012). The supertree was time-calibrated using the stratigraphic age of each taxon (as mean of time interval) (see Tables S3 and S5). The assignment of branch lengths was performed in R (R Development Core Team, 2011) using the APE package (version 2.7-2; Paradis, Claude & Strimmer, 2004) and a protocol written by Graeme Lloyd (see http://www.graemetlloyd.com/methdpf.html) for adjusting zero branch lengths by sharing out the time equally between branches (see Brusatte et al., 2008; Brusatte, 2011), and adding an arbitrary length of 1 million years to the root. The time-calibrated supertree was imported into the software package Mesquite 2.72 (Maddison & Maddison, 2009). Subsequently, Procrustes coordinates and centroid sizes of the adult representatives of the taxa were mapped onto the supertree as continuous characters using square change parsimony. This algorithm performs an ancestral state reconstruction by collating the sum of squared changes of continuous characters along all branches of a tree and estimates the most parsimonious ancestral states by minimizing the total sum of squared changes across the tree (Maddison, 1991). In the next step we tested if the continuous data contains a phylogenetic signal. We performed a permutation test in MorphoJ in which the topology was held constant and both the Procrustes-fitted shape data and the centroid size for each taxon were randomly permuted for all the terminals across the tree 10,000 times (Laurin, 2004; Klingenberg & Gidaszewski, 2010). The data are considered to contain a statistically significant phylogenetic signal if the squared length of the original supertree occurs in at least 95% of the randomly generated trees. Additionally, we quantified phylogenetic signal in our data using a multivariate form of the K statistic with 10,000 replications (Blomberg, Garland & AR, 2003; Paradis, 2012; Adams, 2014) in R using the package geomorph (Adams & Otárola-Castillo, 2013). This test estimates the strength of a phylogenetic signal in a data set in relation to a simulated Brownian motion model, which is expressed as K and p values. To obtain ancestral ontogenetic trajectories, the protocol described above was repeated in a new nexus file containing the Procrustes-fitted shapes and centroid sizes of the juvenile

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 7/41 specimens. As the juvenile data set is only represented by five taxa, the original supertree was pruned such that only these taxa remained, retaining the original time-calibration. Finally, the ancestral Procrustes-fitted shapes and centroid sizes of both juvenile and adult Saurischia, Neotheropoda, Orionides and Avetheropoda were exported and combined with the respective data from the ontogenetic trajectories of the terminal taxa. The ancestral Procrustes-fitted shape of Averostra was not considered because no ceratosaur juveniles have been published in detail so far (see Madsen & Welles, 2000). The new data set including the five terminal and four ancestral ontogenetic trajectories was loaded again into MorphoJ. Regression analyses of ontogenetic trajectories A multivariate, pooled within-group regression of shape against log-transformed centroid size including terminal taxa and hypothetical ancestors (see above) was performed (Piras et al., 2011; Bhullar et al., 2012; Zelditch, Swiderski & Sheets, 2012), in which the Procrustes coordinates were transformed into a regression score (see Drake & Klingenberg, 2008). In contrast to many previous studies of heterochrony using geometric morphometrics, which compare only the ontogenetic trajectories of terminal taxa, our approach allows the determination of possible heterochronic patterns between ancestors and descendants. The different ontogenetic trajectories were compared regarding slope, length, angles and range of shape variation spanned by the predicted regression score. The angles between ontogenetic trajectories were calculated based on Procrustes distances and centroid sizes (see above). As mentioned above, studies of heterochrony require size, shape and ontogenetic age as independent vectors (Klingenberg, 1998). Due to missing data on the individual age of the specimens, ontogenetic age could not be taken into account. As a consequence, the regression analysis explores allometry and not heterochrony (Klingenberg & Spence, 1993; Klingenberg, 1998; Gould, 2000). While some heterochronic processes can result from allometric changes (e.g., acceleration and neoteny), allometric studies allow only conclusions regarding paedomorphosis and peramorphosis (Klingenberg & Spence, 1993; Klingenberg, 1998), which are expressed by the shape vector (i.e., regression score). Peramorphosis can be inferred if the adult individual of the descendant trajectory falls along higher regression scores than the respective ancestral one, whereas paedomorphosis can be inferred based along lower scores. To verify the results of such regression analyses we repeated the analysis using Euclidean distance, which is equivalent to Procrustes distance (see Singleton, 2002; Tallman et al., 2013) as a separate shape vector measuring differences in shape. The Euclidean distance matrix was calculated in PAST 3.05 (Hammer, Harper & Ryan, 2001) on the basis of the Procrustes coordinates of terminal taxa and hypothetical ancestors (see above), which were exported from MorphoJ. For regression analysis, the juvenile specimen of Massospondylus, which represents the sample with the smallest centroid size, was set to zero for aligning the distance values of the remaining taxa (Fig. 4). To test if the shape changes, and as a result the presence of heterochrony, of an ancestor- descendant relationship are statistically meaningful, we calculated the confidence interval (CI) of the differences between regression scores and Euclidean distances of terminal and ancestral taxa (n = 68) and compared them with the differences of ancestral and descendant

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 8/41 regression scores from the sub-sample containing the ontogenetic trajectories. Changes were considered significant if the differences between regression scores were at least 1.5 times higher than the CI value (see Cumming, Fidler & Vaux, 2007). For comparison, we performed another PCA with the data set containing just terminal and ancestral ontogenetic trajectories and calculated the angles and lengths of the trajectories on the basis of the first two principal components, which were found to contain all significant shape information based on the broken stick method (see above). Finally, the ancestral shape reconstructions calculated for the adult representatives of the taxa were used to qualitatively discuss the evolutionary changes within basal Sauropodomorpha and Theropoda with respect to the ontogenetic changes and heterochronic trends found in the different trajectories.

RESULTS General ontogenetic changes The juveniles of the sauropodomorph Massospondylus and the theropods that were sampled here tend to have skulls with a short and abruptly tapering snout, short antorbital fenestrae, large subcircular orbits, slender jugals, and dorsoventrally deep orbital and postorbital regions relative to the snout. In addition, the jaw joint is more anteriorly placed relative to the occiput, with exception of the juvenile specimen of Allosaurus sampled here. The general ontogenetic pattern includes an elongated and dorsoventrally deeper snout relative to the orbital and postorbital regions, and also a relative increase in size of the antorbital fenestra, which correlates with a relative decrease in size of the orbit. Finally, the jugal becomes more massive in all taxa, which is more pronounced in the large-bodied theropods Allosaurus and Tarbosaurus (Fig. 1). The relative elongation of the snout and antorbital fenestra were not observed in the Allosaurus or Tarbosaurus ontogenies, which is probably due to the fact that the juveniles sampled do not represent the earliest ontogenetic stages (Loewen, 2009; Tsuihiji et al., 2011, see ‘Discussion’). However, the discovery of an isolated maxilla identified as a hatchling allosauroid might indicate that the snout of early Allosaurus juveniles was probably short and subsequently increased in relative length during early ontogeny (Rauhut & Fechner, 2005). In addition to these more general ontogenetic modifications, individual taxa show specific shape changes (Fig. 1): (a) In Massospondylus the external naris becomes larger and expands dorsally. The postorbital also becomes relatively more robust. The infratemporal fenestra decreases in relative size. The jaw joint moves anteroventrally. (b) In Coelophysis the external naris becomes smaller and shifts anteriorly. The notch of the alveolar margin between the premaxilla and maxilla decreases in relative size during ontogeny, while the alveolar margin of the premaxilla becomes more aligned with that of the maxilla. The descending process of the lacrimal becomes more slender anteroposteriorly. The postorbital becomes more gracile in its relative shape. The infratemporal fenestra increases in relative size. The jaw joint moves posterodorsally. (c) In the megalosaurid taxon, the external naris becomes relatively larger and expands posteriorly. The lacrimal is inclined strongly backwards and the postorbital becomes

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 9/41 Figure 1 Ontogenetic changes in the skull of saurischian dinosaurs. (A) General ontogenetic pattern in Saurischia exemplified for the basal theropod Coelophysis bauri (adult specimen modified after Rauhut, 2003). (B–F) Specific ontogenetic changes in saurischian dinosaurs visualized as wireframes of Procrustes- fitted shapes. (B) Massospondylus. (C) Coelophysis bauri. (D) Megalosaurid taxon. (E) Allosaurus. (F) Tar- bosaurus. Grey dashed lines represent the juvenile stage and black solid lines represent the adult stage.

relatively more robust. The infratemporal fenestra increases in its relative size. The jaw joint moves posteriorly. (d) In Allosaurus the external naris does not change in relative size, but shifts ventrally. The descending process of the lacrimal becomes more massive anteroposteriorly. The lacrimal develops a prominent dorsal horn through ontogeny. In contrast to previous taxa, the postorbital region of Allosaurus increases dorsoventrally such that the postorbital, quadratojugal and squamosal become relatively more robust. The ventral shift of the jugal leads to the formation of a wide angle between the ventral margins of the maxilla and jugal. Due to its posteroventral expansion, the postorbital affects the shape of the infratemporal fenestra. However, the infratemporal fenestra does not decrease in its relative size, but shifts anteroventrally. The jaw joint moves anteroventrally.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 10/41 (e) In Tarbosaurus the external naris does not change in relative size, but shifts dorsally. As in Allosaurus, the descending process of the lacrimal becomes more massive. The same is true for the postorbital region, which increases in depth dorsoventrally. This change is correlated with the development of a more robust postorbital, quadratojugal and squamosal. The jaw joint moves posteroventrally. Principal component analysis and phylogenetic correlation The first three principal components account for 68.0% of the total variation (PC 1: 30.8%; PC 2: 23.9%; PC 3: 13.3%), in which PC 2 and PC 3 contain the main allometric shape information (see Table S12). PC 1 describes the overall skull depth, size and anteroposterior position of the external naris, length of the premaxilla, size of the maxillary antorbital fossa, and position of the lacrimal and postorbital on the anteroposterior axis (affecting the size of the antorbital fenestra, orbit and infratemporal fenestra). The dorsoventral dimension of the orbit is affected by the relative depth of the entire orbital and postorbital regions, while that of the infratemporal fenestra is affected by the relative position of the jugal-quadratojugal bar. The variation in the depth of the skull also affects the position of the jaw joint on the dorsoventral axis (Fig. 2C). PC 2 describes the length of the snout caused by variation in the length of the maxilla and inclination and anteroposterior position of the lacrimal. The inclination of the lacrimal affects the size of the antorbital fenestra, while both position and inclination affect the anteroposterior dimension of the orbit. PC 2 also accounts for the length and the dorsoventral position of the external naris and size of the upper temporal region (Fig. 2C). PC 3 describes the length of the premaxilla, posterior extension of the external naris, dorsoventral height of the maxilla, and anteroposterior dimension of the ventral process of the lacrimal (which affects the shape of the antorbital fenestra and orbit). The shape of the orbit is further affected by the anteroposterior dimension of the jugal-postorbital bar. Further variation captured by PC 3 is related to the shape of the skull roof in the orbital and postorbital regions, dorsoventral height of the infratemporal fenestra, and position of the jaw joint on the anterodorsal-posteroventral axis (Fig. 2C). The permutation tests and the multivariate K statistic recovered that both Procrustes- fitted shapes (tree length weighted by branch lengths = 0.5108, p < 0.0001; K = 0.2607, p = 0.0016) and centroid size (tree length weighted by branch lengths = 8.3598, p = 0.0005; K = 0.8900, p = 0.0002) are correlated with phylogeny. Furthermore, the multivariate regression analysis reveals that skull shape is significantly correlated with centroid size (correlation index: 15.32%, p < 0.0001) (Fig. 4A, Table S12). Ontogenetic trajectories in the PCA morphospace Based on the PCA results of the original data set (i.e., including semi-landmarks), the ontogenetic trajectories are not uniform (Fig. 2 and Table 1). The trajectory of Allosaurus is short and mainly explained by shape variation captured by PC 1, while that of Tarbosaurus is also short, but mainly explained by PCs 1 and 3. The third principal component has stronger influence on the ontogenetic shape variation in Tarbosaurus based on the length of its trajectory. Compared to Allosaurus and Tarbosaurus, the other ontogenetic trajectories

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 11/41 Figure 2 Principal component analysis of the main sample. (A) Ontogenetic trajectories of terminal taxa for PC 1 versus PC 2. (B) Ontogenetic trajectories of terminal taxa for PC 1 against PC 3. (C) Illustra- tion of the main shape changes for the first three principal components. Theropod taxa are shown as black dots, while sauropodomorph taxa are shown as grey dots. The arrows illustrate the different ontogenetic trajectories, in which the arrowhead marks the position of the adult individual.

are longer. The trajectory of Coelophysis is mainly explained by the shape variation captured by PCs 1 and 2, while its slope is opposite to the direction along PC 1 compared to the trajectories of Allosaurus and Tarbosaurus. Based on the angles, the ontogenetic trajectories of Massospondylus and the megalosaurid taxon are mainly influenced by the shape variation captured by PCs 2 and 3, in which the ontogenetic trajectory of Massospondylus is directed

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 12/41 Table 1 Angles and length of terminal ontogenetic trajectories. Angles of ontogenetic trajectories against PC 1, pairwise angles between ontogenetic trajectories in the PC 1–PC 2 and PC 1–PC 3 morphospace and length of ontogenetic trajectories in the PC 1–PC 2 and PC 1–PC 3 morphospace (Figs. 2A and 2B). Green fields mark pairwise angles in the PC 1–PC 2 morphospace and orange fields mark that of the PC 1–PC 3 morphospace. Angles, lengths and slopes of ontogenetic trajectories versus log-transformed centroid size (LogCS) (Fig. 4A).

Massospondylus Coelophysis Megalosaurid Allosaurus Tarbosaurus taxon Angle (PC 1–PC 2) 85.6492 42.3458 83.3216 5.3228 3.7406 Length (PC 1–PC 2) 0.1761 0.1174 0.1414 0.0332 0.0403 Angle (PC 1–PC 3) 63.2316 10.1684 65.0464 18.5268 60.5157 Length (PC 1–PC 3) 0.0297 0.0881 0.0390 0.0349 0.0818 Massospondylus – 73.4000 128.2780 98.2417 177.2841 Coelophysis 43.3033 – 54.8780 171.6416 109.3159 Megalosaurid taxon 2.3276 40.9757 – 133.4803 54.4379 Allosaurus 89.0280 132.3313 91.3556 – 79.0425 Tarbosaurus 98.0914 141.3947 100.4190 9.0634 – Angle (LogCS) 3.3947 4.8961 4.6105 0.2535 1.5851 Length (LogCS) 2.2815 1.0636 1.8147 1.0657 1.4016 Slope (LogCS) 0.0593 0.0857 0.0806 0.0044 0.0277

in the opposite direction along PC 3 to that of the megalosaurid taxon and Tarbosaurus. However, the length of the trajectories indicates that the second principal component has major influence on the shape variation in both species during ontogeny. The PCA reveals that the ontogenetic elongation of the snout is primarily related to a relative increase in the length of the maxilla (PCs 1, 2). In Massospondylus and the megalosaurid taxon the ontogenetic elongation of the snout is further affected by the relative increase of the length of the premaxilla (PC 3). The relative increase in snout depth results mainly from a ventral expansion of the maxilla, which is more prominent in Allosaurus and Tarbosaurus than in other taxa (PCs 1, 3). In the megalosaurid taxon and Allosaurus, maxillary deepening occurs together with a dorsoventral expansion of the nasal (PC 1). Additionally, dorsoventral expansion of the premaxilla is observed in Allosaurus and Tarbosaurus (PC 1). The relative elongation of the snout in Massospondylus, the megalosaurid taxon and Coelophysis correlates with a relative increase in the anteroposterior length of the antorbital fenestra, caused by a posterior shift of the lacrimal and elongation of the maxilla (PCs 1, 2). Additionally, in Coelophysis the anterior border of the antorbital fenestra extends anteriorly (PC 1). In both Massospondylus and the megalosaurid taxon, the antorbital fenestra is shifted posteriorly during ontogeny (PC 2). The megalosaurid taxon shows a further dorsal expansion of the antorbital fenestra (PC 3), not seen in the latter two taxa. Although no relative size changes could be observed in the antorbital fenestrae of Allosaurus and Tarbosaurus, the antorbital fenestra of Allosaurus shifts posterodorsally during ontogeny, whereas that of Tarbosaurus shifts ventrally. In most trajectories, the most anterior point of the antorbital fossa shifts posteriorly during ontogeny (PCs 1–3), but a relative decrease in the length of the maxillary antorbital fossa is present in Allosaurus and

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 13/41 Tarbosaurus (PC 1). In the megalosaurid taxon, the anterior margin of the antorbital fossa shifts ventrally, whereas in Coelophysis it shifts anteriorly (PC 1), which correlates with the anterior elongation of the antorbital fenestra in this taxon (see above). As mentioned above, the orbit decreases in relative size in all taxa during ontogeny (PCs 1–3). In Coelophysis and Massospondylus this is related to a relative shift of the lacrimal posteriorly (PCs 1, 2). In the megalosaurid taxon, Allosaurus and Tarbosaurus the relative size reduction is correlated with a change in orbital shape from subcircular to oval. In the megalosaurid taxon these changes are linked to a posterior shift of the lacrimal (PC 2) and anterior shift of the postorbital and ascending process of the jugal (PC 3), which is correlated with an anterior extension of the infratemporal fenestra. In Allosaurus, the ontogenetic changes of the orbit are related to the posterior extension of the lacrimal and anterior shift of the postorbital and ascending process of the jugal (PC 1). Additionally, the orbit of Allosaurus is shifted slightly dorsally. In Tarbosaurus, these changes result from an anterior extension of both the postorbital and ascending process of the jugal (PC 3). The orbit of Tarbosaurus becomes posteriorly constricted by an anterior shift of the ventral process of the postorbital, forming a suborbital process. We examined the differences in the trajectory directions when terminal and ancestral ontogenetic series are compared to each other (Fig. 3 and Table 2). The significant shape variation evaluated via the broken stick method is described by the first two principal components (PC 1: 50.39%; PC 2: 20.79%). Both axes are correlated with centroid size (see Table S12). The ontogenetic trajectory of Coelophysis in mainly influenced by PC 1, while that of the megalosaurid taxon, Massospondylus and all ancestral trajectories is influenced by both PC 1 and 2, in which the first principal component is found to have a higher impact on the shape variation during ontogeny. In contrast, the ontogenetic trajectories of Tarbosaurus and Allosaurus are mainly influenced by PC 2. Ontogenetic trajectories in the regression analyses The ontogenetic trajectory of Massospondylus is longer than that of the hypothetical ancestor of Saurischia for both shape variables (regression score and Euclidean distance), while the values of the shape variables are significantly lower. However, the slope of the trajectory of Massospondylus based on the regression score is less pronounced than that of the saurischian ancestor, while it is more pronounced for the Euclidean distance (Figs. 4B, 4C, Tables 3 and4). In contrast, the ontogenetic trajectory of the hypothetical ancestor of Neotheropoda is slightly longer and has a greater slope, while the regression score and the Euclidean distance of the adult individual are significantly higher than that of the saurischian ancestor. Coelophysis possesses a longer and steeper ontogenetic trajectory for both shape variables with significantly higher values than the hypothetical ancestor of Neotheropoda (Figs. 4B, 4C, Tables 3 and4). The ontogenetic trajectory of the hypothetical ancestor of Orionides is shorter and has a lower slope than that of the neotheropod ancestor. The regression score of the adult individual is significantly higher, while the Euclidean distance is lower, but not significantly different. Compared to the hypothetical ancestor of Orionides, the megalosaurid taxon has a longer and steeper ontogenetic trajectory, with a significantly higher value for both shape variables

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 14/41 Figure 3 Principal component analysis of ontogenetic trajectories. (A) Terminal and ancestral onto- genetic trajectories for PC 1 against PC 2. The arrows illustrate the different ontogenetic trajectories, in which the arrowhead marks the position of the adult individual and the base of the arrow indicates the ju- venile individual. (B) Illustration of the main shape changes for the first two principal components.

(Figs. 4B, 4C, Tables 3 and4). In contrast, the ontogenetic trajectory of the hypothetical ancestor of Avetheropoda is shorter, possessing a lower slope and significantly lower regression score and Euclidean distance for the adult individual, when compared to the hypothetical ancestor of Orionides. The ontogenetic trajectories of Allosaurus and Tarbosaurus are longer than that of the hypothetical ancestor of Avetheropoda. Both trajectories show a slope decrease compared to their common ancestor. Interestingly, the slope is almost zero when the Euclidean distance is applied as shape variable, indicating only minor shape changes during the ontogeny as sampled. For Allosaurus, both shape values of the adult individual are higher than that of the ancestor, but only the regression score is significant. In contrast, the regression score of the adult individual of Tarbosaurus is significantly lower than that of the hypothetical ancestor of Avetheropoda, while the Euclidean distance results in a higher, but non statistically significant value (Figs. 4B, 4C, Tables 3 and4).

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 15/41 ohe l (2016), al. et Foth PeerJ 10.7717/peerj.1589 DOI ,

Table 2 Angles and lengths of terminal and ancestral ontogenetic trajectories. Angles of ontogenetic trajectories against PC 1, pairwise angles between ontogenetic tra- jectories in the PC 1–PC 2 morphospace and length of ontogenetic trajectories in the PC 1–PC 2 morphospace (Fig. 3A).

Saurischia Massospondylus Neotheropoda Coelophysis Orionides Megalosaurid Avetheropoda Allosaurus Tarbosaurus taxon Angle (PC 1–PC 2) 29.5357 15.193 19.6691 4.0256 33.2773 35.5725 29.4664 65.5478 79.1993 Length (PC 1–PC 2) 0.082 0.1372 0.1202 0.1162 0.0879 0.1571 0.0429 0.044 0.0735 Saurischia – Massospondylus 14.3427 – Neotheropoda 9.8666 4.4761 – Coelophysis 25.5101 11.1674 15.6435 – Orionides 3.7416 18.0843 13.6082 29.2517 – Megalosaurid taxon 6.0368 20.3795 15.9034 31.5469 2.2952 – Avetheropoda 0.0693 14.2734 9.7973 25.4408 3.8109 6.1061 – Allosaurus 84.9165 99.2591 94.783 110.4266 81.1749 78.8797 84.9858 – Tarbosaurus 71.265 85.6076 81.1315 96.7751 67.5234 65.2282 71.3343 13.6515 – 16/41 ohe l (2016), al. et Foth PeerJ 10.7717/peerj.1589 DOI ,

Table 3 Angles and lengths of terminal and ancestral ontogenetic trajectories. Angles, lengths and slopes of ontogenetic trajectories from the regression of shape (Re- gression score, RS and Euclidean Distance, ED) versus log-transformed centroid size (LogCS) (Figs. 4B and 4C).

Saurischia Massospondylus Neotheropoda Coelophysis Orionides Megalosaurid taxon Avetheropoda Allosaurus Tarbosaurus Regression (RS) Angle (LogCS) 4.3762 3.8814 5.1181 6.056 4.1743 5.029 3.0083 1.1845 0.7153 Length (LogCS) 1.1084 2.2828 1.3988 1.0657 1.3267 1.8158 0.9628 1.0659 1.4011 Slope (LogCS) 0.0765 0.0678 0.0896 0.1061 0.0730 0.0880 0.0526 0.0207 0.0125 Regression (ED) Angle (LogCS) 3.4145 5.0440 4.0199 5.0905 2.7768 3.4451 1.7014 −0.1758 −0.1087 Length (LogCS) 1.1071 2.2864 1.3967 1.0640 1.3248 1.8121 0.9619 1.0657 1.4010 Slope (LogCS) 0.0597 0.0883 0.0703 0.0891 0.0485 0.0602 0.0297 −0.0031 −0.0019 17/41 Table 4 Overview of heterochronies in saurischian skull shape. The differences of the regression scores (1RS) and the Euclidean distances (1ED) between ancestor-descendent relationships of adult individuals from the regression analysis (Figs. 4B and 4C) and the interpretation regarding heterochrony.

1RS 1ED Heterochrony Saurischia-Massospondylus −0.0262 −0.0446 Paedomorphosis Saurischia-Neotheropoda 0.0629 0.0733 Peramorphosis Neotheropoda-Coelophysis 0.0140 0.0668 Peramorphosis Neotheropoda-Orionides 0.0146 (−0.0079) NA Orionides-megalosaurid taxon 0.0507 0.0497 Peramorphosis Orionides-Avetheropoda −0.0299 −0.0256 Paedomorphosis Avetheropoda-Allosaurus 0.0153 (0.0066) NA Avetheropoda-Tarbosaurus −0.0145 (0.0015) NA 95% CIs 0.0078 0.0098 Significance levels (p = 0.05) 0.0117 0.0147 Notes. 1RS and 1ED values in brackets mark insignificant trends. NA, not available.

Based on the regression analysis, taxa with higher regressions scores tend to have elongated skulls with long and slender snouts that have a rounded anterior end, and possess anteroposteriorly long antorbital fenestrae, oval orbits and a post-rostrum only slightly dorsoventrally higher than the snout. The maxilla increases in its relative length, but also expands ventrally. The ascending process of the maxilla, the anterior and ascending processes of the jugal, and postorbital become more massive. In contrast, low regression scores account for skull shapes where these features are less pronounced, developed or even show opposite trends. When compared to the regression analyses containing all taxa, the relative position, length and slopes of the ontogenetic trajectories of the terminal taxa is almost identical (Fig. 4A), supporting the robustness of the results recovered.

DISCUSSION Ontogenetic patterns Our knowledge of the cranial ontogeny of non-avian dinosaurs remains fragmentary. Previous studies on cranial ontogeny have often been based on single species (Gow, Kitching & Raath, 1990; Carr & Williamson, 2004; Horner & Goodwin, 2006; Hübner & Rauhut, 2010; Campione & Evans, 2011; Mallon et al., 2011; Canale et al., 2014; Frederickson & Tumarkin-Deratzian, 2014), while only a small number of studies have investigated this topic on the interspecific level (Carr, 1999; Evans, 2010; Bhullar et al., 2012; Mallon, Ryan & Campbell, 2015). As is common in other animal groups, closely related species often undergo similar ontogenetic changes (see Evans, 2010; Mallon, Ryan & Campbell, 2015), while ontogenetic trajectories become more different with increased phylogenetic distance (see Bhullar et al., 2012) or in the case of a single taxon evolving extreme ontogenies compared to their relatives (Horner & Goodwin, 2009, see also Erickson et al., 2004). Despite the large phylogenetic distance between the ontogenetic series sampled here, the present study reveals that the cranial ontogeny of saurischian dinosaurs undergoes some general patterns, including the relative elongation and dorsoventrally heightening of the preorbital region, decrease in orbit size and increase in jugal robustness. However, the PCA shows that the different ontogenetic trajectories differ strongly in length, direction

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 18/41 Figure 4 Centroid size regression analyses for the main sample. (A) Regression analysis of all terminal taxa including ontogenetic trajectories against log-transformed skull centroid size (LogCS) (p < 0.0001). (B) Regression analysis of only terminal (solid arrows) and ancestral (dashed arrows) ontogenetic trajecto- ries against log centroid size (p < 0.0001) using the regression score as shape variable. (C) Equivalent re- gression analysis to (B) using the Euclidean distance as shape variable. Theropod taxa are shown as black dots, while sauropodomorph taxa are shown as grey dots. The arrows illustrate the different ontogenetic trajectories, in which the arrowhead marks the position of the adult individual and the base of the arrow indicates the juvenile individual.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 19/41 and also the location within the morphospace. Here, the theropod taxa are markedly separated from the sauropodomorph Massospondylus, which is morphologically very distinct from other basal sauropod taxa. This is especially obvious in the large distance within morphospace between Massospondylus and Coelophysis, which represent the most basal ontogenetic series of each clade indicating a strong diversification of skull shape in the early evolution of Saurischia. This may be related to differentiations along both lines of Saurischia in terms of ecology, including trophic specializations (see Tykoski & Rowe, 2004; Barrett & Rayfield, 2006; Langer et al., 2010; Sakamoto, 2010). The fact that the separation within the morphospace already take place among juvenile specimens indicates that these specializations might appear very early in ontogenetic development. Although the distances among such specimens in morphospace are large, the trajectories show that both species still share similar trends in cranial development (Figs. 2 and3). Although occupying a similar area of morphospace, the ontogenetic trajectory of the megalosaurid taxon differs markedly from that of Allosaurus and Tarbosaurus, showing more similarity with that of Massospondylus and Coelophysis, which share in common the relative elongation of the snout. The latter process probably represents a morphological trend within megalosaurids (Therrien & Henderson, 2007; Sadleir, Barrett & Powell, 2008), while large-bodied allosauroids and tyrannosaurids tend to have rather deeper than long skulls (see Brusatte et al., 2012; Foth & Rauhut, 2013a). However, as several medium-sized tyrannosauroids also have elongated snouts (Li et al., 2010; Brusatte, Carr & Norell, 2012; Lü et al., 2014; Porfiri et al., 2014), their ontogenetic trajectories would probably more closely resemble that of the megalosaurid taxon. One has to take into account that the length and direction of the ontogenetic trajectories of Allosaurus and Tarbosaurus are likely influenced by the fact that the juvenile specimens are ontogenetically more developed compared to that of the megalosaurid taxon (see below). Assuming that the hatchlings of Allosaurus and Tarbosaurus also had short, tapering snouts, the trajectory would probably be more similar in length and direction to that of the megalosaurid taxon. Heterochronic patterns Previous workers have hypothesized that skull shape diversity in theropods and sauropodomorphs was driven by phylogenetic interrelationships, dietary preferences (Young & Larvan, 2010; Brusatte et al., 2012; Foth & Rauhut, 2013a), functional constraints (Henderson, 2002; Foth & Rauhut, 2013a), but also heterochrony (Long & McNamara, 1997; Bhullar et al., 2012). This study builds on the recent heterochronic analysis of Bhullar et al. (2012), who primarily examined derived non-avian theropods and basal avians on the basis of a great number of ontogenetic trajectories of non-avian coelurosaurs and an extant phylogenetic bracket of crocodylians and birds, covering a broader scale of archosaurian craniofacial shape variation. However, by sampling and comparing ontogenetic trajectories of more basal saurischian taxa, our data set allows for reevaluation of the conclusions presented by Bhullar et al. (2012) with regards to basal sauropodomorphs, allosauroids and tyrannosauroids. The current study supports the influence of heterochrony on the cranial evolution of some saurischian lineages. When the differences of the regressions scores (1RS) and the Euclidean distances (1ED) in an ancestor-descendant relationship are compared,

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 20/41 the significant decrease of the shape values indicates potential paedomorphosis for the skull shape of Massospondylus and the hypothetical ancestor of Avetheropoda, while the skulls of Coelophysis, the megalosaurid taxon and the hypothetical ancestor of Neotheropoda, might be peramorphic. Thus, the current analyses support a paedomorphosis for basal sauropodomorphs as predicted by Bhullar et al. (2012). Due to contradicting results regarding shape differences, no heterochronic pattern can be inferred for Allosaurus, Tarbosaurus and the hypothetical ancestor of Orionides. Thus, the current analyses do not support the predicted cranial peramorphosis for the allosauroids and tyrannosaurid lineage (Long & McNamara, 1997; Bhullar et al., 2012; Canale et al., 2014), while studies on growth (Bybee, Lee & Lamm, 2006; Erickson et al., 2004) and body size evolution (Dececchi & Larsson, 2013; Benson et al., 2014; Lee et al., 2014) in theropods indicate such a trend. However, this conflict is probably caused by incomplete sampling of ontogenetic trajectories, which affects the estimated shape of the hypothetical ancestor of Avetheropoda. A further expansion of the sampling of ontogenetic trajectories of saurischian taxa and the inclusion of an extant phylogenetic bracket (see Bhullar et al., 2012), would probably change some aspects of the analytical outcomes of this study (see below). The increase in slopes in the ontogenetic trajectories of Neotheropoda, Coelophysis and the megalosaurid taxon, when compared to their ancestors, might show evidence for peramorphic acceleration. However, with a few exceptions, bone histology of basal theropods (e.g., Coelophysis and Syntarsus) is not well studied, so that this cannot be confirmed by growth patterns. Several studies on body size evolution support a peramorphic trend, showing an increase of size from the hypothetical ancestor of Saurischia over Neotheropoda towards megalosaurids (Irmis, 2011; Dececchi & Larsson, 2013). In contrast, the relative decrease in slope in the ontogenetic trajectory of the hypothetical ancestor of Avetheropoda might indicate neoteny. But again this cannot be confirmed by bone histological data at this time. Dececchi & Larsson (2013) and Lee et al. (2014) found a decrease of body size from the hypothetical ancestor of Tetanurae towards Avetheropoda, supporting a paedomorphic trend in body size. For Massospondylus, the situation is not entirely clear, as our two shape variables led to conflicting results regarding the slope, when compared with the saurischian ancestor. Thus, no underlying heterochronic process can be diagnosed for the paedomorphic skull shape of Massospondylus. Although basal sauropodomorphs show a gradual trend towards bigger body size (Sander et al., 2010; Irmis, 2011; Benson et al., 2014) and longer, accelerated growth (Chinsamy, 1993; Erickson, Rogers & Yerby, 2001; Klein & Sander, 2007), skull size decreased relatively (Rauhut et al., 2011). This relative shrinking might be the reason for the maintenance of a more juvenile skull shape in the early evolution of sauropodomorphs. However, due to the lack of information regarding the ontogenetic age of the individuals, the deduction of heterochronic process related to the slope (i.e., neoteny and acceleration) has to be considered with caution (see below). The results of the regression analyses can be further used to interpret evolutionary shape changes found between hypothetical ancestors and terminal taxa in the ancestral shape reconstruction analyses of the main sample (i.e., continuous character mapping of the Procrustes-fitted shapes) in terms of paedomorphic or peramorphic trends (Fig. 5).

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 21/41 Figure 5 Simplified phylogeny of Saurischia showing the main heterochronic trends of the skull. Peramorphosis is colored in green and paedomorphosis in yellow. Grey trends indicate uncertain shape trends. Shape of the hypothetical ancestors based on the continuous character mapping of the Procrustes- fitted shapes of the adult terminal taxa from the original data set. Blue skulls represent ancestral skull shapes for which ontogeny could not be analysed. The heterochronic trends found in the regression analyses are visualized by the color of the branches. Possible heterochronic trends related to the skull evolution of allosauroids and basal coelurosaurs (see ‘Discussion’) are shown as dashed branches.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 22/41 Comparing the skull shape of the hypothetical ancestor of Saurischia to that of Sauropodomorpha indicates a possible initial paedomorphosis in the evolution of the latter group as shown by the regression analyses, which is depicted by a decrease in the relative length of the preorbital region and an increase in the relative orbital size and depth of the postorbital region. As stated above, Bhullar et al. (2012) already proposed a cranial paedomorphosis for basal sauropodomorphs after finding a strong similarity between the skull shape of Eoraptor and the juvenile theropod Coelophysis, which had been also highlighted qualitatively by previous authors (e.g., Ezcurra, 2007). In addition, Foth (2013) has shown that the skull shape of Eoraptor and Pampadromaeus resembled that of the juvenile theropods Sciurumimus and Juravenator. In Eusauropoda the snouts become more aberrant due to a dorsal shift of the external naris, posterodorsal extension of the premaxilla, elongation of the ascending process of the maxilla and modification of the postorbital region, affecting the relative size of the jugal and postorbital, which become more gracile (Wilson & Sereno, 1998; Rauhut et al., 2011). While the shape changes in the snout and the shift of the naris were previously presumed to be peramorphic (Long & McNamara, 1997), one can assume on the basis of the current observations that the increase of gracility in the postorbital region of derived sauropods may result from modular paedomorphosis. In this context, Salgado (1999) has hypothesized that the reduction of the supratemporal fenestra and fusion of the frontals in diplodocoid sauropods is the result of a peramorphic heterochrony, while the loss of contact between squamosal and quadratojugal could be paedomorphic. However, these character changes are beyond the scope of the current study due to the lack of good skull material of juvenile individuals of basal sauropods, and thus, need to be analysed in more detail in future studies after the appropriate juvenile materials are discovered. In contrast, the initial evolutionary changes in the skull shape of Theropoda were driven by peramorphic events, as is observed in Coelophysis, the megalosaurid taxon and the hypothetical ancestor of Neotheropoda in the regression analyses. These changes include the elongation of the snout, increase in length of the antorbital fenestra, and trends to a relatively smaller orbit and more robust post-rostral region. The basal ceratosaur Limusaurus has a rather small skull with a short snout, enlarged subcircular orbit and gracile jugal and postorbital, so it is possible that the more robust skull shape (oval orbit, massive jugal and postorbital) of large-bodied ceratosaurs like Ceratosaurus and abelisaurids (e.g., Carnotaurus and Majungasaurus) could be the result of a secondary peramorphosis as it was proposed for allosaurids and tyrannosaurids (e.g., Long & McNamara, 1997; Bhullar et al., 2012). However, due to the poor cranial knowledge and fluctuating phylogenetic relationships of basal ceratosaurs from the Early and Middle Jurassic (e.g., Pol & Rauhut, 2012; Tortosa et al., 2013), the early skull shape evolution of Ceratosauria is not currently reproducible. In contrast, the skull of the hypothetical ancestor of Avetheropoda is probably paedomorphic with respect to that of Orionides as shown in the regression analyses (Figs. 4B and 4C). This trend might extend to the hypothetical ancestor of Coelurosauria, Maniraptoriformes and Maniraptora, leading to a shorter, more tapering snout in lateral view, smaller antorbital fenestrae, enlarged subcircular orbits, and a more gracile postrostral region, resembling the skull shape of the juvenile megalosaurid Sciurumimus. These

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 23/41 findings may indicate that the paedomorphic trend hypothesized for Eumaniraptora by Bhullar et al. (2012) reaches back into the early evolution of Avetheropoda, and that basal coelurosaurs in fact represent ‘‘miniaturized’’ tetanurans, conserving juvenile characters in adult individuals. A similar trend is found for body size evolution in theropods, showing a successive decrease in body size within Avetheropoda (Novas et al., 2012; Dececchi & Larsson, 2013; Lee et al., 2014). In contrast to this supposed early paedomorphic trend, the ancestral shape reconstruction reveals that the skulls of allosauroids become secondarily more robust in relation to the hypothetical ancestor of Avetheropoda, supporting cranial peramorphosis (see Canale et al., 2014). This might also be the case for large-bodied tyrannosaurids (see Long & McNamara, 1997; Bhullar et al., 2012), although the current regression analyses could not find such a signal for both groups (see below). Bhullar et al. (2012) suggested a multi-step progenetic paedomorphosis for skull shape of Paraves and basal birds, with modular peramorphic trends related to beak formation, and further peramorphic trends for secondarily large-bodied troodontids and dromaeosaurids. These heterochronic changes were supported by trends regarding body size evolution (Turner et al., 2007; Dececchi & Larsson, 2013; Lee et al., 2014) and growth patterns (Erickson et al., 2009) found within Eumaniraptora. However, as it is the case for Sauropodomorpha, various trends seen in skull shape evolution of theropods need to be verified in the future regarding possible heterochrony on the basis of new material of both juvenile and adult specimens. Functional and ecological implications The major differences in cranial shape found here clearly affect dietary preferences and functional constraints. The robust morphology of the postorbital region and the oval orbit in peramorphic skulls was previously discussed in relation to the generation of higher bite forces (Henderson, 2002; Foth & Rauhut, 2013a). However, these functional constraints go hand in hand with a decrease in cranial disparity (Brusatte et al., 2012). Paedomorphic changes in the orbital and postorbital regions were discussed in relation to visual elaboration and brain enlargement (Bhullar et al., 2012), and may have played an important role in nocturnal activity (Schmitz & Motani, 2011) or the evolution of flight within Paraves (Balanoff et al., 2013). On the other hand, large and circular orbits might simply correlate with reduced mechanical stresses during biting (Henderson, 2002), which have been suggested to also influence size and shape of the external naris, antorbital fenestra and infratemporal fenestra (Witmer, 1997; Witzel & Preuschoft, 2005; Witzel et al., 2011). Both ontogenetic and phylogenetic variations in snout shape are likely related to dietary preferences (Brusatte et al., 2012; Foth & Rauhut, 2013a; Foth, Bona & Desojo, 2015; see above), in which the shape of premaxillae and maxillae partly determines the number and size of teeth (Henderson & Weishampel, 2002). Various examples of ontogenetic changes in the morphology and number of teeth are documented in Saurischia, including the basal sauropodomorph Massospondylus, coelophysoids (Colbert, 1989), basal tetanurans (Rauhut & Fechner, 2005; Rauhut et al., 2012), tyrannosaurids (Carr, 1999; Tsuihiji et al., 2011) and maniraptorans (Kundrát et al., 2008; Bever & Norell, 2009). Based on these observations the evolutionary increase in the number of teeth has been interpreted as peramorphic

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 24/41 (Bever & Norell, 2009). Tooth morphology, however, was found to be a stronger indicator of diet than the shape of the snout itself (see Smith, 1993; Barrett, 2000; Barrett, Butler & Nesbitt, 2011; Zanno & Makovicky, 2011; Foth & Rauhut, 2013a; Hendrickx & Mateus, 2014). In this context, Rauhut et al. (2012) hypothesised based on the similarities in the dentition of the juvenile megalosaurid Sciurumimus, adult compsognathids (Stromer, 1934; Currie & Chen, 2001; Peyer, 2006) and adult dromaeosaurids (Xu & Wu, 2001; Norell et al., 2006), that strongly recurved crowns with reduced or no mesial serrations may be paedomorphic in the latter two taxa. This heterochrony probably results from the decrease of body size observed in coelurosaurs (see above) and indicates an evolutionary shift in dietary preferences to smaller prey (see also Zanno & Makovicky, 2011). Limitations As is common in vertebrate paleontology, the current study has a limited sample size when compared with extant neontological data sets (Brown & Vavrek, 2015). The current results are necessarily preliminary and must be viewed with caution especially because the sampling of ontogenetic trajectories is considerably lower than the sampling of adult individuals. Furthermore, trajectories are constructed using a single juvenile and adult specimen, with no intermediate forms. A single multistage example for Tyrannosaurus presented by Bhullar et al. (2012) has shown that during ontogeny the trajectory can change its direction considerably in a multivariate PCA plot. This, in turn, has an important impact on the length of the trajectory and its angle in relation to other trajectories. However, in regression analyses the difference with a two-stage approach should be less substantial as multivariate shape information is transformed into a single variable of shape for each stage with respect to its centroid size. The poor sample of juveniles is a result of rarity and poor preservation in the fossil record, which seems to be due to a number of factors, including preferred hunting of juveniles by predators (Hone & Rauhut, 2010) and a smaller likelihood of preservation, discovery, and collection because juveniles have smaller body sizes and more fragile bones than adults (Brown et al., 2013). Thus, due to small sample sizes, the statistical power of our analyses is generally low (see Cumming, Fidler & Vaux, 2007), limiting the explanatory power of our results. On the other hand, Brown & Vavrek (2015) recently demonstrated that the number of positive and negative allometries is underestimated in smaller samples in both paleontological and neontological data sets. Another issue affecting our results is that the juvenile individuals sampled here are all of different early ontogenetic stages. The juvenile Massospondylus represents a composite of several embryos close to hatching (Reisz et al., 2010); the megalosaurid taxon (i.e., Sciurumimus albersdoerferi) is an early juvenile and its exact age could not be determined (Rauhut et al., 2012); the age of the Coelophysis juvenile reconstructed is approximately one year old (estimated by Colbert, 1990; Rinehart et al., 2009); the juvenile Tarbosaurus specimen is two to three years old (Tsuihiji et al., 2011); and the juvenile Allosaurus is likely five to seven years old (estimated based on Bybee, Lee & Lamm, 2006; Loewen, 2009). Thus, the different ontogenetic stages of the juvenile specimens and the small number of individuals for each ontogenetic series most likely affected the length, but maybe also the slope of the calculated trajectories (and thus the angles between the trajectories)

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 25/41 (see Cardini & Elton, 2007), including that of the hypothetical ancestors. Furthermore, the uncertainty regarding the age of the specimens leads to another weak point, as specimen age was not used to characterize the ontogenetic trajectories (see above), which is a common problem in paleontology (e.g., McKinney, 1986; Klingenberg, 1998; Gould, 2000; Schoch, 2010; Bhullar et al., 2012). In consequence, the applied regression analyses explored allometry and not heterochrony (see Klingenberg & Spence, 1993; Klingenberg, 1998). The substitution of age by size, however, would imply similar growth dynamics (i.e., proportionality between age and size) between ancestors and descendants, which would consequently ignore heterochronic processes related to growth rates (i.e., progenesis and acceleration). Although dinosaurs generally have higher growth rates compared to other non-avian reptiles, histological studies reveal that growth rates are not identical (Erickson, Rogers & Yerby, 2001; Erickson et al., 2004; Padian, De Ricqlès & Horner, 2001; Sander et al., 2004; Erickson et al., 2009; Grady et al., 2014; Werner & Griebeler, 2014). Therefore, allometric patterns cannot be used to infer heterochrony beyond paedomorphosis and peramorphosis as argued by Klingenberg & Spence (1993) and Klingenberg (1998). Taking the uncertainties related to the lengths and slopes of the ontogenetic trajectories (due to incomplete ontogenetic series) and statistical uncertainties (due to the small sample size) into account, the classifications of underlying heterochronic processes would be misleading and probably erroneous. In the current study, the interpretations of paedomorphosis and peramorphosis rely on the significant shape differences between adult individuals of the ontogenetic trajectories expressed by shape vectors in the regression analyses, for which the multivariate shape data were transformed into a univariate shape variable. These differences are affected by type of shape variable, but more importantly by the ancestral shapes, which in turn depend on the phylogenetic relationships, the algorithm of time calibration (e.g., Bapst, 2014) and the method of reconstruction (e.g., Martins, 1999; Webster & Purvis, 2002). Thus, one has to be aware that the application of different methods could result in slightly different ancestral shapes, affecting the value of the shape variable. However, because the current sample covers all major linages of basal saurischians except of crested taxa, which were found to impact the ancestral shape of the skull roofs significantly (see Fig. S5 and Table S6), the results of the ancestral reconstruction of adult individuals are viewed as valid. By using two different shape variables (Regression score and Euclidean distance), it was possible to confirm significant results through multiple methods. The undefined trend found for Tarbosaurus in relation to the hypothetical ancestor of Avetheropoda illustrates the limitations of our analyses. Our result is seemingly contradictory to previous hypotheses and our ancestral shape reconstruction, which proposed peramorphosis as the main driver of skull evolution in large-bodied tyrannosaurids (see above, Long & McNamara, 1997; Bhullar et al., 2012). As stated above, this result is most likely related to the small sample size of ontogenetic trajectories as skulls with elongated and slender snouts are considered to be peramorphic on the basis of the regression analyses. The inclusion of more ontogenetic trajectories of large- bodied theropods would probably change this result in favour of a trend towards a deeper snout. Furthermore, large-bodied tyrannosaurids like Tarbosaurus descended

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 26/41 from small-bodied coelurosaurian ancestors (Xu et al., 2004; Xu et al., 2006; Brusatte et al., 2010; Rauhut, Milner & Moore-Fay, 2010; Benson et al., 2014), which means that the hypothetical inclusion of an ontogenetic trajectory of a small-bodied basal coelurosaur (e.g., Compsognathus, Dilong, Haplocheirus) and a respective hypothetical ancestor of Coelurosauria would probably change the current results, leading to a secondary peramorphic trend in Late Cretaceous tyrannosaurids, as suggested by previous authors. Thus, this result is very likely an artefact of incomplete sampling. In this context, the limited number of ontogenetic series of basal sauropodomorphs results only in a rough trend regarding the relationship between cranial ontogeny and evolution, which cannot be extended to more general patterns in the skull shape evolution of basal sauropods.

CONCLUSIONS The importance of heterochrony in non-avian dinosaur skull evolution is a relatively new concept (see Long & McNamara, 1997; Bhullar et al., 2012). This study quantitatively assesses the impact of skull heterochrony across early saurischian evolution, allowing testing some of the heterochronic trends proposed by Bhullar et al. (2012) and further highlights different vantages of using morphometric data to elucidate heterochronic trends. We estimated hypothetical ontogenetic trajectories in Saurischia, Neotheropoda, Orionides, and Avetheropoda using ontogenetic trajectories of Massospondylus, Coelophysis, a megalosaurid taxon, Allosaurus and Tarbosaurus. When compared using PCA, the ontogenetic trajectories of the terminal taxa show great variation in length and direction, but still follow some very general patterns, including a relatively elongated and dorsoventrally deeper preorbital region, decrease in orbit size and increase in jugal robustness. General peramorphic skulls include more elongate and slender snouts, elongate antorbital fenestrae, oval orbits, dorsoventrally shallower post-rostral regions, and more massive maxillae, jugals, and postorbitals. Paedomorphic skulls show the opposite features. The shape changes from the hypothetical ancestor of Saurischia to Massospondylus were paedomorphic, as previously suggested by Bhullar et al. (2012). In contrast, skull evolution of basal theropod taxa was probably affected by peramorphic trends. However, Avetheropoda showed paedomorphic changes compared to Orionides. This might indicate that the paedomorphic trend found for Eumaniraptora (see Bhullar et al., 2012) may reach back into the early evolution of Avetheropoda. The hypothesized peramorphic evolution for skull shape of allosaurids and tyrannosaurids could not be supported by the current study, but this probably resulted from the small sample size of ontogenetic trajectories. Although our data showed minimal differences between our crested-taxa and non-crested taxa data sets and semi-landmark and no semi-landmark data sets, it is important to fully evaluate all possible sources of trends, especially when working with a small data set. As stated above, our study is hampered by the preservation of the fossil record (mainly the poor sample of complete juvenile specimens) and more finds will help to elucidate other evolutionary patterns related to heterochrony. With a larger number of taxa comprising juvenile and adult stages it will be possible to further test heterochronic hypotheses within Saurischia in more detail, and eliminate artefacts related to sample size. Future studies may also examine ontogenetic

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 27/41 histories of individual taxa that have reasonably complete ontogenetic samples, such as Coelophysis, to evaluate which factors (dietary preference, heterochrony, etc.) drive shape change in individual taxa. A larger number of studies using geometric morphometrics for individual taxa as well as a more complete sampling within Saurischia are necessary to more completely assess the importance of heterochronic processes in both sauropodomorph and theropod skull shape. In addition, it would be of value to explore modularity in saurischian skulls to project the investigation of heterochronic processes to particular skull regions. In sum, this study demonstrates that heterochrony played an important role in basal non-avian saurischian skull evolution building upon previous studies (Bhullar et al., 2012). Institutional Abbreviations

BMMS Bürgermeister Müller Museum Solnhofen, Solnhofen, Germany CM Carnegie Museum of Natural History, Pittsburgh, USA GR Ruth Hall Museum, Ghost Ranch, USA IVPP Institute of Vertebrate Paleontology and Paleoanthropology, Beijing, China MCZ Museum of Comparative Zoology, Harvard University, USA

ACKNOWLEDGEMENTS We thank Oliver Rauhut (Bayerische Staatssammlung für Paläontologie und Geologie, München), Miriam Zelditch (University of Michigan), Johannes Knebel (Ludwig Maximilians University, München), Stefan Richter (University of Rostock), Walter Joyce and Eduardo Ascarrunz (both University of Fribourg) for discussion, and Michel Laurin (Sorbonne Universités, Paris) for comments on an earlier version of the manuscript. We further thank Matthew Lamanna (Carnegie Museum of Natural History, Pittsburgh), Alex Downs (Ruth Hall Museum, Ghost Ranch), David Gillette (Museum of Northern Arizona, Flagstaff) and Xu Xing (Institute of Vertebrate Paleontology and Paleoanthropology, Beijing) for access to collections. This study benefitted especially from critical comments of Jesús Marugán-Lobón (Universidad Autónoma de Madrid) and three anonymous reviewers.

ADDITIONAL INFORMATION AND DECLARATIONS

Funding CF is supported by a DFG grant to Oliver Rauhut (RA 1012/12-1) and a postdoctoral fellowship of the DAAD German Academic Exchange Service (No. 9154678), BPH is supported by a Benjamin Franklin Fellowship at the University of Pennsylvania, MDE is supported by a grant of the DFG Emmy Noether Programme to Richard J. Butler (BU 2587/3-1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 28/41 Grant Disclosures The following grant information was disclosed by the authors: DFG: RA 1012/12-1. DAAD German Academic Exchange Service: 9154678. Benjamin Franklin Fellowship. DFG Emmy Noether Programme: BU2587/3-1. Competing Interests The authors declare there are no competing interests. Author Contributions • Christian Foth conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, wrote the paper, prepared figures and/or tables, reviewed drafts of the paper. • Brandon P. Hedrick and Martin D. Ezcurra conceived and designed the experiments, contributed reagents/materials/analysis tools, wrote the paper, reviewed drafts of the paper. Data Availability The following information was supplied regarding data availability: All raw data are submitted as Supplemental Information, including a MorphoJ and Nexus file. Supplemental Information Supplemental information for this article can be found online at http://dx.doi.org/10.7717/ peerj.1589#supplemental-information.

REFERENCES Adams DC. 2014. A generalized K statistic for estimating phylogenetic signal from shape and other high-dimensional multivariate data. Systematic Biology 63:685–697 DOI 10.1093/sysbio/syu030. Adams DC, Otárola-Castillo E. 2013. geomorph: an R package for the collection and analysis of geometricmorphometric shape data. Methods in Ecology and Evolution 4:393–399 DOI 10.1111/2041-210X.12035. Adams DC, Rohlf FJ, Slice DE. 2004. Geometric morphometrics: ten years of progress following the ‘‘revolution’’. Italian Journal of Zoology 71:5–16 DOI 10.1080/11250000409356545. Adams DC, Rohlf FJ, Slice DE. 2013. A field comes of age: geometric morphometrics in the 21st century. Hystrix 24:7–14. Alberch P, Gould SJ, Oster GF, Wake DB. 1979. Size and shape in ontogeny and phylogeny. Paleobiology 5:296–317. Balanoff AM, Bever GS, Rowe TB, Norell MA. 2013. Evolutionary origins of the avian brain. Nature 201:93–96.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 29/41 Balanoff AM, Rowe TB. 2007. Osteological description of an embryonic skeleton of the extinct elephant bird, Aepyornis (Palaeognathae: Ratitae). Journal of Vertebrate Paleontology 27:1–53. Bapst DW. 2014. Assessing the effect of time-scaling methods on phylogeny-based analyses in the fossil record. Paleobiology 40:331–351 DOI 10.1666/13033. Barrett PM. 2000. Prosauropod dinosaurs and iguanas: speculations on the diets of extinct reptiles. In: Sues H-D, ed. Evolution of herbivory in terrestrial vertebrates. Cambridge: Cambridge University Press, 42–78. Barrett PM, Butler RJ, Nesbitt SJ. 2011. The roles of herbivory and omnivory in early dinosaur evolution. Earth and Environmental Science Transactions of the Royal Society of Edinburgh 101:383–396. Barrett PM, Rayfield EJ. 2006. Ecological and evolutionary implications of dinosaur feeding behaviour. Trends in Ecology and Evolution 21:217–224 DOI 10.1016/j.tree.2006.01.002. Benson RBJ, Campione NE, Carrano MT, Mannion PD, Sullivan C, Upchurch P, Evans SE. 2014. Rates of dinosaur body mass evolution indicate 170 million years of sustained ecological innovation on the avian stem lineage. PLoS Biology 12:e1001853 DOI 10.1371/journal.pbio.1001853. Berge C, Penin X. 2004. Ontogenetic allometry, heterochrony, and interspecific differ- ences in the skull of african apes, using tridimensional procrustes analysis. American Journal of Physical Anthropology 124:124–138 DOI 10.1002/ajpa.10333. Bever GS, Norell MA. 2009. The perinate skull of Byronosaurus (Troodontidae) with observations on the cranial Ontogeny of paravian theropods. American Museum Novitates 3657:1–51 DOI 10.1206/650.1. Bhullar B-A. 2012. A phylogenetic approach to ontogeny and heterochrony in the fossil record: cranial evolution and development in anguimorphan lizards (Reptilia: Squamata). Journal of Experimental Zoology Part B, Molecular and Developmental Evolution 318B:521–530. Bhullar B-A, Marugán-Lobón J, Racimo F, Bever GS, Rowe TB, Norell MA, Abzhanov A. 2012. Birds have paedomorphic dinosaur skulls. Nature 487:223–226 DOI 10.1038/nature11146. Blomberg S, Garland TJ, Ives AR. 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57:717–745 DOI 10.1111/j.0014-3820.2003.tb00285.x. Bonnan MF. 2004. Morphometric analysis of humerus and femur shape in Morrison sauropods: implications for functional morphology and paleobiology. Paleobiology 30:444–470 DOI 10.1666/0094-8373(2004)030<0444:MAOHAF>2.0.CO;2. Bookstein FL. 1991. Morphometric tools for landmark data. Cambridge: Cambridge University Press. Bookstein FL, Schäfer K, Prossinger H, Seidler H, Fiedler M, Stringer C, Weber GW, Arsuaga J-L, Slice DE, Rohlf FJ, Recheis W, Mariam AJ, Marcus LF. 1999. Comparing frontal cranial profiles in archaic and modern Homo by morphometric

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 30/41 analysis. The Anatomical Record 257:217–224 DOI 10.1002/(SICI)1097-0185(19991215)257:6<217::AID-AR7>3.0.CO;2-W. Brown CM, Evans DC, Campione NE, O’Brien LJ, Eberth DA. 2013. Evidence for tapho- nomic size bias in the Dinosaur Park Formation (Campanian, Alberta), a model Mesozoic terrestrial alluvial-paralic system. Palaeogeography, Palaeoclimatology, Palaeoecology 372:108–122 DOI 10.1016/j.palaeo.2012.06.027. Brown CM, Vavrek MJ. 2015. Small sample sizes in the study of ontogenetic allometry; implications for palaeobiology. PeerJ 3:e818 DOI 10.7717/peerj.818. Brusatte SL. 2011. Calculating the tempo of morphological evolution: rates of discrete character change in a phylogenetic context. In: Elewa AMT, ed. Computational Paleontology. Heidelberg: Springer, 53–74. Brusatte SL, Benton MJ, Ruta M, Lloyd GT. 2008. Superiority, competition, and opportunism in the evolutionary radiation of dinosaurs. Science 321:1485–1488 DOI 10.1126/science.1161833. Brusatte SL, Carr TD, Norell MA. 2012. The osteology of Alioramus, a gracile and long-snouted tyrannosaurid (Dinosauria: Theropoda) from the Late Cretaceous of Mongolia. Bulletin of the American Museum of Natural History 366:1–197 DOI 10.1206/770.1. Brusatte SL, Montanari S, Sakamoto M, Harcourt-Smith WEH. 2012. The evolution of cranial form and function in theropod dinosaurs: insight from geometric morphometrics. Journal of Evolutionary Biology 25:365–377 DOI 10.1111/j.1420-9101.2011.02427.x. Brusatte SL, Norell MA, Carr TD, Erickson GM, Hutchinson JR, Balanoff AM, Bever GS, Choiniere JN, Makovicky PJ, Xu X. 2010. Tyrannosaur paleobiology: new research on ancient exemplar organisms. Science 329:1481–1485 DOI 10.1126/science.1193304. Butler RJ, Goswami A. 2008. Body size evolution in Mesozoic birds: little evidence for Cope’s rule. Journal of Evolutionary Biology 21:1673–1682 DOI 10.1111/j.1420-9101.2008.01594.x. Bybee PJ, Lee AH, Lamm E-T. 2006. Sizing the Jurassic theropod dinosaur Allosaurus: assessing growth strategy and evolution of ontogenetic scaling of limbs. Journal of Morphology 267:347–359 DOI 10.1002/jmor.10406. Cabreira SF, Schultz CL, Bittencourt JS, Soares MB, Fortier DC, Silva LR, Langer MC. 2011. New stem-sauropodomorph (Dinosauria, Saurischia) from the Triassic of Brazil. Naturwissenschaften 98:1035–1040 DOI 10.1007/s00114-011-0858-0. Campione NE, Evans DC. 2011. Cranial growth and variation in edmontosaurs (Di- nosauria: Hadrosauridae): implications for Latest Cretaceous megaherbivore diver- sity in North America. PLoS ONE 6:e25186 DOI 10.1371/journal.pone.0025186. Canale IC, Novas FE, Salgado L, Coria RA. 2014. Cranial ontogenetic variation in Mapusaurus roseae (Dinosauria: Theropoda) and the probable role of heterochrony in carcharodontosaurid evolution. Paläontolologische Zeitschrift 1–11.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 31/41 Cardini A, Elton S. 2007. Sample size and sampling error in geometric morphometric studies of size and shape. Zoomorphology 126:121–134 DOI 10.1007/s00435-007-0036-2. Carr TD. 1999. Craniofacial ontogeny in Tyrannosauridae (Dinosauria, Coelurosauria). Journal of Vertebrate Paleontology 19:497–520 DOI 10.1080/02724634.1999.10011161. Carr TD, Williamson TE. 2004. Diversity of late Maastrichtian Tyrannosauridae (Dinosauria: Theropoda) from western North America. Zoological Journal of the Linnean Society 142:479–523 DOI 10.1111/j.1096-3642.2004.00130.x. Carrano MT, Benson RBJ, Sampson SD. 2012. The phylogeny of Tetanurae (Dinosauria: Theropoda). Journal of Systematic Palaeontology 10:211–300 DOI 10.1080/14772019.2011.630927. Chinnery B. 2004. Morphometric analysis of evolutionary trends in the ceratopsian postcranial skeleton. Journal of Vertebrate Paleontology 24:591–609 DOI 10.1671/0272-4634(2004)024[0591:MAOETI]2.0.CO;2. Chinsamy A. 1993. Bone histology and growth trajectory of the prosauropod dinosaur Massospondylus carinatus Owen. Modern Geology 18:319–329. Colbert EH. 1989. The Triassic dinosaur Coelophysis. Museum of Northern Arizona Bulletin 57:1–160. Colbert EH. 1990. Variation in Coelophysis bauri. In: Carpenter K, Currie PJ, eds. Dinosaur systematics: approaches and perspectives. Cambridge: Cambridge University Press, 81–90. Collyer ML, Adams DC. 2007. Analysis of two-state multivariate phenotypic change in ecological studies. Ecology 88:683–692 DOI 10.1890/06-0727. Corti M. 1993. Geometric morphometrics: an extension of the revolution. Trends in Ecology and Evolution 8:302–303. Cumming G, Fidler F, Vaux DL. 2007. Error bars in experimental biology. The Journal of Cell Biology 177:7–11 DOI 10.1083/jcb.200611141. Currie PJ, Chen P. 2001. Anatomy of Sinosauropteryx prima from Liaoning, northeastern China. Canadian Journal of Earth Sciences 38:1705–1727 DOI 10.1139/e01-050. Dal Sasso C, Maganuco S. 2011. Scipionyx samniticus (Theropoda: Compsognathidae) from the Lower Cretaceous of Italy. Memorie della Società Italiana di Scienze Naturali e del Museo Civico di Storia Naturale di Milano 37:1–281. Dececchi TA, Larsson HCE. 2013. Body and limb size dissociation at the origin of birds: uncoupling allometric constraints across a macroevolutionary transition. Evolution 67:2741–2752 DOI 10.1111/evo.12150. Drake AG. 2011. Dispelling dog dogma: an investigation of heterochrony in dogs using 3D geometric morphometric analysis of skull shape. Evolution & Development 13:204–213 DOI 10.1111/j.1525-142X.2011.00470.x. Drake AG, Klingenberg CP. 2008. The pace of morphological change: historical transformation of skull shape in St Bernard dogs. Proceedings of the Royal Society B 275:71–76 DOI 10.1098/rspb.2007.1169.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 32/41 Erickson GM, Makovicky PJ, Currie PJ, Norell MA, Yerby SA, Brochu CA. 2004. Gigantism and comparative life-history parameters of tyrannosaurid dinosaurs. Nature 430:772–775 DOI 10.1038/nature02699. Erickson GM, Rauhut OWM, Zhou Z, Turner AH, Inouye BD, Hu D, Norell MA. 2009. Was dinosaurian physiology inherited by birds? Reconciling slow growth in Archaeopteryx. PLoS ONE 4:e7390 DOI 10.1371/journal.pone.0007390. Erickson GM, Rogers KC, Yerby SA. 2001. Dinosaur growth patterns and rapid avian growth rates. Nature 412:429–432 DOI 10.1038/35086558. Evans DC. 2010. Cranial anatomy and systematics of Hypacrosaurus altispinus, and a comparative analysis of skull growth in lambeosaurine hadrosaurids (Di- nosauria: Ornithischia). Zoological Journal of the Linnean Society 159:398–434 DOI 10.1111/j.1096-3642.2009.00611.x. Ezcurra MD. 2007. The cranial anatomy of the coelophysoid theropod Zupaysaurus rougieri (Upper Triassic, Argentina). Historical Biology 19:185–202 DOI 10.1080/08912960600861467. Ezcurra MD. 2012. Phylogenetic analysis of Late Triassic-Early Jurassic neotheropod dinosaurs: implications for the early theropod radiation [Abstract]. Journal of Vertebrate Paleontology, Program and Abstracts 32:91. Ezcurra MD, Butler RJ. 2015. Post-hatchling cranial ontogeny in the Early Triassic diapsid reptile Proterosuchus fergusi. Journal of Anatomy 226:387–402. Ezcurra MD, Novas FE. 2005. Phylogenetic relationships of the Triassic theropod Zupaysaurus rougieri from NW Argentina [Abstract]. II Congresso Latino-Americano de Paleontologia de Vertebrados 102–104. Ezcurra MD, Novas FE. 2007. Phylogenetic relationships of the Triassic theropod Zupaysaurus rougieri from NWArgentina. Historical Biology 19:35–72 DOI 10.1080/08912960600845791. Fink WL. 1982. The conceptual relationship between ontogeny and phylogeny. Paleobiology 8:254–264. Forasiepi AM, Sánchez-Villagra MR. 2014. Heterochrony, dental ontogenetic diversity, and the circumvention of constraints in marsupial mammals and extinct relatives. Paleobiology 40:222–237 DOI 10.1666/13034. Foth C. 2013. Ontogenetic, macroevolutionary and morphofunctional patterns in archosaur skulls: a morphometric approach. München: Ludwig-Maximilians-Universität. Foth C, Bona P, Desojo JB. 2015. Intraspecific variation in the skull morphology of the black caiman Melanosuchus niger (Alligatoridae, Caimaninae). Acta Zoologica (Stockholm) 96:1–13. Foth C, Rauhut OWM. 2013a. Macroevolutionary and morphofunctional patterns in theropod skulls: a morphometric approach. Acta Palaeontologica Polonica 58:1–16. Foth C, Rauhut OWM. 2013b. The good, the bad, and the ugly: the influence of skull reconstructions and intraspecific variability in studies of cranial morphometrics in theropods and basal saurischians. PLoS ONE 8:e72007 DOI 10.1371/journal.pone.0072007.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 33/41 Frederickson JA, Tumarkin-Deratzian AR. 2014. Craniofacial ontogeny in Centrosaurus apertus. PeerJ 2:e252 DOI 10.7717/peerj.252. Fritsch M, Bininda-Emonds ORP, Richter S. 2013. Unraveling the origin of Cladocera by identifying heterochrony in the developmental sequences of Branchiopoda. Frontiers in Zoology 10:35 DOI 10.1186/1742-9994-10-35. Gerber S, Neige P, Eble GJ. 2007. Combining ontogenetic and evolutionary scales of morphological disparity: a study of early Jurassic ammonites. Evolution & Development 9:472–482 DOI 10.1111/j.1525-142X.2007.00185.x. Gould SJ. 1977. Ontogeny and phylogeny. Cambridge: Harvard University Press. Gould SJ. 2000. Of coiled oysters and big brains: how to rescue the terminol- ogy of heterochrony, now gone astray. Evolution & Development 2:241–248 DOI 10.1046/j.1525-142x.2000.00067.x. Gow CE, Kitching JW, Raath MA. 1990. Skulls of the prosauropod dinosaur Mas- sospondylus carinatus Owen in the collections of the Bernand Price Institute for Palaeontological Research. Palaeontologia Africana 27:45–58. Gower JC. 1975. Generalized Procrustes analysis. Psychometrika 40:33–51 DOI 10.1007/BF02291478. Grady JM, Enquist BJ, Dettweiler-Robinson E, Wright NA, Smith FA. 2014. Evidence for mesothermy in dinosaurs. Science 344:1268–1272 DOI 10.1126/science.1253143. Guenther MF. 2009. Influence of sequence heterochrony on hadrosaurid di- nosaur postcranial development. The Anatomical Record 292:1427–1441 DOI 10.1002/ar.20988. Hammer O, Harper DAT, Ryan PD. 2001. PAST: paleontological statistics software package for education and data analysis. Palaeontologia Electronica 4:1–9. Hedrick BP, Dodson P. 2013. Lujiatun psittacosaurids: understanding individual and taphonomic variation using 3D geometric morphometrics. PLoS ONE 8:e69265 DOI 10.1371/journal.pone.0069265. Henderson DM. 2002. The eyes have it: the sizes, shapes, and orientations of theropod orbits as indicators of skull strength and bite force. Journal of Vertebrate Paleontology 22:766–778. Henderson DM, Weishampel DB. 2002. Convergent evolution of the maxilla-dental- complex among carnivorous archosaurs. Senckenbergiana Lethaea 82:77–92 DOI 10.1007/BF03043774. Hendrickx C, Mateus O. 2014. Abelisauridae (Dinosauria: Theropoda) from the Late Jurassic of Portugal and dentition-based phylogeny as a contribution for the identification of isolated theropod teeth. Zootaxa 3759:1–74 DOI 10.11646/zootaxa.3759.1.1. Hennig W. 1966. Phylogenetic systematics. Urbana: University of Illinois Press. Hone DWE, Naish D, Cuthill I. 2012. Does mutual sexual selection explain the evolution of head crests in pterosaurs and dinosaurs? Lethaia 45:139–156 DOI 10.1111/j.1502-3931.2011.00300.x. Hone DWE, Rauhut OWM. 2010. Feeding behaviour and bone utilization by theropod dinosaurs. Lethaia 43:232–244.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 34/41 Horner JR, Goodwin MB. 2006. Major cranial changes during Triceratops ontogeny. Proceedings of the Royal Society B 273:2757–2761 DOI 10.1098/rspb.2006.3643. Horner JR, Goodwin MB. 2009. Extreme cranial ontogeny in the Upper Cretaceous di- nosaur Pachycephalosaurus. PLoS ONE 4:e7626 DOI 10.1371/journal.pone.0007626. Hübner TR, Rauhut OWM. 2010. A juvenile skull of Dysalotosaurus lettowvorbecki (Ornithischia: Iguanodontia), and implications for cranial ontogeny, phylogeny, and taxonomy in ornithopod dinosaurs. Zoological Journal of the Linnean Society 160:366–396 DOI 10.1111/j.1096-3642.2010.00620.x. Irmis RB. 2011. Evaluating hypotheses for the early diversification of dinosaurs. Earth and Environmental Science Transactions of the Royal Society of Edinburgh 101:397–426. Jackson DA. 1993. Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches. Ecology 74:2204–2214 DOI 10.2307/1939574. Klein N, Sander PM. 2007. Bone histology and growth of the prosauropod dinosaur Plateosaurus engelhardti von Mayer, 1837 from the Norian bonebeds of Trossingen (Germany) and Frick (Switzerland). Special Papers in Palaeontology 77:169–206. Klingenberg CP. 1998. Heterochrony and allometry: the analysis of evolutionary change in ontogeny. Biological Reviews 73:79–123 DOI 10.1017/S000632319800512X. Klingenberg CP. 2011. MorphoJ: an integrated software package for geometric morpho- metrics. Molecular Ecology Resources 11:353–357 DOI 10.1111/j.1755-0998.2010.02924.x. Klingenberg CP, Gidaszewski NA. 2010. Testing and quantifying phylogenetic signals and homoplasy in morphometric data. Systematic Biology 59:245–261 DOI 10.1093/sysbio/syp106. Klingenberg CP, Spence JR. 1993. Heterochrony and allometry: lessons from the water strider genus Limnoporus. Evolution 47:1834–1853 DOI 10.2307/2410225. Kundrát M, Cruickshank ARI, Manning TW, Nudds J. 2008. Embryos of ther- izinosauroid theropods from the Upper Cretaceous of China: diagnosis and analysis of ossification patterns. Acta Zoologica (Stockholm) 89:231–251. Langer MC, Ezcurra MD, Bittencourt JS, Novas FE. 2010. The origin and early evolution of dinosaurs. Biological Reviews 85:55–110 DOI 10.1111/j.1469-185X.2009.00094.x. Laurin M. 2004. The evolution of body size, Cope’s rule and the origin of amniotes. Systematic Biology 53:594–622 DOI 10.1080/10635150490445706. Lautenschlager S. 2014. Morphological and functional diversity in therizinosaur claws and the implications for theropod claw evolution. Proceedings of the Royal Society B 281:20140497 DOI 10.1098/rspb.2014.0497. Lee MSY, Cau A, Naish D, Dyke GJ. 2014. Sustained miniaturization and anatom- ical innovation in the dinosaurian ancestors of birds. Science 345:562–566 DOI 10.1126/science.1252243. Li D, Norell MA, Gao K, Smith ND, Makovicky PJ. 2010. A longirostrine tyrannosauroid from the Early Cretaceous of China. Proceedings of the Royal Society B 277:183–190 DOI 10.1098/rspb.2009.0249.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 35/41 Lieberman DE, Carlo J, Ponc de Leóon M, Zollikofer CPE. 2007. A geometric morpho- metric analysis of heterochrony in the cranium of chimpanzees and bonobos. Journal of Human Evolution 52:647–662 DOI 10.1016/j.jhevol.2006.12.005. Loewen MA. 2009. Variation in the Late Jurassic theropod dinosaur Allosaurus: ontogenetic, functional, and taxonomic implications. Salt Lake City: University of Utah. Loewen MA, Irmis RB, Sertich JJW, Currie PJ, Sampson SD. 2013. Tyrant dinosaur evolution tracks the rise and fall of Late Cretaceous oceans. PLoS ONE 8:e79420 DOI 10.1371/journal.pone.0079420. Long JA, McNamara KJ. 1997. Heterochrony: the key to dinosaur evolution. In: Wolberg DL, Stumps E, Rosenberg GD, eds. Dinofest international. Philadelphia: Academy of Natural Sciences, 113–123. Lü J, Yi L, Brusatte SL, Yang L, Li H, Chen L. 2014. A new clade of Asian Late Cretaceous long-snouted tyrannosaurids. Nature Communications 5:3788. Maddison WP. 1991. Squared-change parsimony reconstructions of ancestral states for continuous-valued characters on a phylogenetic tree. Systematic Zoology 40:304–314 DOI 10.2307/2992324. Maddison WP, Maddison DR. 2009. Mesquite: a modular system of evolutionary analysis. Version 3.02. Madsen JHJ, Welles SP. 2000. Ceratosaurus (Dinosauria, Theropoda), a revised osteol- ogy. Utah Geology Survey Miscellaneous Publication 00-2:1–80. Maiorini L, Farke AA, Kotsakis T, Piras P. 2015. Males resemble females: re-evaluating sexual dimorphism in Protoceratops andrewsi (Neoceratopsia, Protoceratopsidae). PLoS ONE 10:e0126464 DOI 10.1371/journal.pone.0126464. Mallon JC, Holmes R, Eberth DA, Ryan MJ, Anderson JS. 2011. Variation in the skull of Anchiceratops (Dinosauria, Ceratopsidae) from the Horseshoe Canyon Formation (Upper Cretaceous) of Alberta. Journal of Vertebrate Paleontology 31:1047–1071 DOI 10.1080/02724634.2011.601484. Mallon JC, Ryan MJ, Campbell JA. 2015. Skull ontogeny in Arrhinoceratops brachyops (Ornithischia: Ceratopsidae) and other horned dinosaurs. Zoological Journal of the Linnean Society 175:910–929. Martínez RN. 2009. Adeopapposaurus mognai, gen. et sp. nov. (Dinosauria: Sauropodomor- pha), with comments on adaptations of basal Sauropodomorpha. Journal of Vertebrate Paleontology 29:142–164 DOI 10.1671/039.029.0102. Martínez RN, Apaldetti C, Abelin D. 2013. Basal sauropodomorphs from the Is- chigualasto Formation. Society of Vertebrate Paleontology Memoir 12:51–69. Martínez RN, Sereno PC, Alcober OA, Colombi CE, Renne PR, Montañez IP, Currie BS. 2011. A basal dinosaur from the dawn of the dinosaur era in southwestern Pangaea. Science 331:206–210 DOI 10.1126/science.1198467. Martins EP. 1999. Estimation of ancestral states of continuous characters: a computer simulation study. Systematic Biology 48:642–650 DOI 10.1080/106351599260210. McKinney ML. 1986. Ecological causation of heterochrony: a test and implications for evolutionary theory. Paleobiology 12:282–289. McNamara KJ. 1982. Heterochrony and phylogenetic trends. Paleobiology 8:130–142.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 36/41 McNamara KJ, McKinney ML. 2005. Heterochrony, disparity, and macroevolution. Paleobiology 31:17–26 DOI 10.1666/0094-8373(2005)031[0017:HDAM]2.0.CO;2. Mitteroecker P, Gunz P. 2009. Advances in geometric morphometrics. Evolutionary Biology 36:235–247 DOI 10.1007/s11692-009-9055-x. Mitteroecker P, Gunz P, Bookstein FL. 2005. Heterochrony and geometric morphomet- rics: a comparison of cranial growth in Pan paniscus versus Pan troglodytes. Evolution & Development 7:244–258 DOI 10.1111/j.1525-142X.2005.05027.x. Mitteroecker P, Gunz P, Weber GW, Bookstein FL. 2004. Regional dissociated hete- rochrony in multivariate analysis. Annals of Anatomy 186:463–470 DOI 10.1016/S0940-9602(04)80085-2. Molnar RE. 2005. Sexual selection and sexual dimorphism in theropods. In: Carpenter K, ed. The carnivorous dinosaurs. Bloomington: Indiana University Press, 284–312. Norell MA, Clark JM, Turner AH, Makovicky PJ, Barsbold R, Rowe TB. 2006. A new droameosaurid theropod from Ukhaa Tolgod (Ömnögov, Mongolia). American Museum Novitates 3545:1–51 DOI 10.1206/0003-0082(2006)3545[1:ANDTFU]2.0.CO;2. Novas FE, Ezcurra MD, Agnolín FL, Pol D, Ortíz R. 2012. New Patagonian Cretaceous theropod sheds light about the early radiation of Coelurosauria. Revista del Museo Argentino de Ciencias Naturales, nueva serie 14:57–81. Padian K, De Ricqlès AJ, Horner JR. 2001. Dinosaurian growth rates and bird origins. Nature 412:405–408 DOI 10.1038/35086500. Padian K, Horner JR. 2011. The evolution of ‘‘bizarre structures’’ in dinosaurs: biome- chanics, sexual selection, social selection or species recognition? Journal of Zoology 283:3–17 DOI 10.1111/j.1469-7998.2010.00719.x. Paradis E. 2012. Analysis of phylogenetics and evolution with R. New York: Springer. Paradis E, Claude J, Strimmer K. 2004. APE: analyses of phylogenetics and evolution in R language. Bioinformatics 20:289–290 DOI 10.1093/bioinformatics/btg412. Peyer K. 2006. A reconsideration of Compsognathus from the Upper Tithonian of Canjuers, southeastern France. Journal of Vertebrate Paleontology 26:879–896 DOI 10.1671/0272-4634(2006)26[879:AROCFT]2.0.CO;2. Piras P, Salvi D, Ferrara G, Maiorino L, Delfino M, Pedde L, Kotsakis T. 2011. The role of post-natal ontogeny in the evolution of phenotypic diversity in Podarcis lizards. Journal of Evolutionary Biology 24:2705–2720 DOI 10.1111/j.1420-9101.2011.02396.x. Pol D, Rauhut OWM. 2012. A Middle Jurassic abelisaurid from Patagonia and the early diversification of theropod dinosaurs. Proceedings of the Royal Society B 279:3170–3175 DOI 10.1098/rspb.2012.0660. Porfiri JD, Novas FE, Calvo JO, Agnolín FL, Ezcurra MD, Cerda IA. 2014. Juvenile specimen of Megaraptor (Dinosauria, Theropoda) sheds light about tyrannosauroid radiation. Cretaceous Research 51:35–55 DOI 10.1016/j.cretres.2014.04.007. R Development Core Team. 2011. R: a language and environment for statistical computing. Raff RA. 1996. The shape of life: genes, development, and the evolution of animal form. Chicago: University of Chicago Press.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 37/41 Rauhut OWM. 2003. The interrelationships and evolution of basal theropod dinosaurs. Special Papers in Palaeontology 69:1–213. Rauhut OWM, Fechner R. 2005. Early development of the facial region in a non- avian theropod dinosaur. Proceedings of the Royal Society B 272:1179–1183 DOI 10.1098/rspb.2005.3071. Rauhut OWM, Fechner R, Remes K, Reis K. 2011. How to get big in the Mesozoic: the evolution of the sauropodomorph body plan. In: Klein N, Remes K, Gee CT, Sander PM, eds. Biology of the sauropod dinosaurs: understanding the life of giants. Bloomington: Indiana University Press, 119–149. Rauhut OWM, Foth C, Tischlinger H, Norell MA. 2012. Exceptionally preserved juvenile megalosauroid theropod dinosaur with filamentous integument from the Late Jurassic of Germany. Proceedings of the National Academy of Sciences of the United States of America 109:11746–11751 DOI 10.1073/pnas.1203238109. Rauhut OWM, Milner AC, Moore-Fay S. 2010. Cranial osteology and phylogenetic position of the theropod dinosaur Proceratosaurus bradleyi (Woodward, 1910) from the Middle Jurassic of England. Zoological Journal of the Linnean Society 158:155–195 DOI 10.1111/j.1096-3642.2009.00591.x. Reilly SM, Wiley EO, Meinhardt DJ. 1997. An integrative approach to heterochrony: the distinction between interspecific and intraspecific phenomena. Biological Journal of the Linnean Society 60:119–143 DOI 10.1111/j.1095-8312.1997.tb01487.x. Reisz RR, Evans DC, Sues H-D, Scott D. 2010. Embryonic skeletal anatomy of the sauropodomorph dinosaur Massospondylus from the Lower Jurassic of South Africa. Journal of Vertebrate Paleontology 30:1653–1665 DOI 10.1080/02724634.2010.521604. Rinehart LF, Lucas SG, Heckert AB, Spielmann JA, Celeskey MD. 2009. The paleobi- ology of Coelophysis bauri (Cope) from the Upper Triassic (Apachean) Whitaker quarry, New Mexico, with detailed analysis of a single quarry block. New Mexico Museum of Natural History and Science 45:1–260. Rohlf FJ. 2005. tpsDig, digitize landmarks and outlines. Version 2.05. Rohlf FJ, Marcus LF. 1993. A revolution in morphometrics. Trends in Ecology and Evolution 8:129–132 DOI 10.1016/0169-5347(93)90024-J. Rohlf FJ, Slice DE. 1990. Extensions of the Procrustes method for the optimal superim- position of landmarks. Systematic Zoology 39:40–59 DOI 10.2307/2992207. Rowe TB. 1989. A new species of the theropod dinosaur Syntarsus from the Early Jurassic Kayenta Formation of Arizona. Journal of Vertebrate Paleontology 9:125–136 DOI 10.1080/02724634.1989.10011748. Sadleir RW, Barrett PM, Powell HP. 2008. The anatomy and systematics of Eustrep- tospondylus oxoniensis, a theropod dinosaur from the Middle Jurassic of Oxfordshire, England. Monograph of the Palaeontological Society 627:1–82. Sakamoto M. 2010. Jaw biomechanics and the evolution of biting performance in theropod dinosaurs. Proceedings of the Royal Society B 277:3327–3333 DOI 10.1098/rspb.2010.0794. Salgado L. 1999. The macroevolution of the Diplodocimorpha (Dinosauria; Sauropoda): a developmental model. Ameghiniana 36:203–216.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 38/41 Sampson SD. 1999. Sex and destiny: the role of mating signals in speciation and macroevolution. Historical Biology 13:173–197 DOI 10.1080/08912969909386580. Sander PM, Christian A, Clauss M, Fechner R, Gee CT, Griebeler EM, Gunga H-C, Hummel J, Mallison H, Perry SF, Preuschoft H, Rauhut OWM, Remes K, Tüttken T, Wings O, Witzel U. 2010. Biology of the sauropod dinosaurs: the evolution of gigantism. Biological Reviews 86:117–155. Sander PM, Klein N, Buffetaut E, Cuny G, Suteethorn V, Le Loeuff J. 2004. Adaptive radiation in sauropod dinosaurs: bone histology indicates rapid evolution of giant body size through acceleration. Organisms, Diversity & Evolution 4:165–173 DOI 10.1016/j.ode.2003.12.002. Schmitz L, Motani R. 2011. Nocturnality in dinosaurs inferred from scleral ring and orbit morphology. Science 332:705–708 DOI 10.1126/science.1200043. Schoch RR. 2009. Life-cycle evolution as response to diverse lake habitats in Paleozoic amphibians. Evolution 63:2738–2749 DOI 10.1111/j.1558-5646.2009.00769.x. Schoch RR. 2010. Heterochrony: the interplay between development and ecol- ogy exemplified by a Paleozoic amphibian clade. Paleobiology 36:318–334 DOI 10.1666/09011.1. Schoch RR. 2014. Amphibian skull evolution: the developmental and functional context of simplification, bone loss and heterotopy. Journal of Experimental Zoology Part B, Molecular and Developmental Evolution 322B:619–630. Schwarz-Wings D, Böhm N. 2014. A morphometric approach to the specific separation of the humeri and femora of Dicraeosaurus from the Late Jurassic of Tendaguru/Tan- zania. Acta Palaeontologica Polonica 59:81–98. Singleton M. 2002. Patterns of cranial shape variation in the Papionini (Primates: Cerco- pithecinae). Journal of Human Evolution 42:547–578 DOI 10.1006/jhev.2001.0539. Slice DE. 2007. Geometric morphometrics. Annual Review of Anthropology 36:261–281 DOI 10.1146/annurev.anthro.34.081804.120613. Smith KK. 1993. The form of the feeding apparatus in terrestrial vertebrates: studies of adaptation and constraint. In: Hanken J, Hall BK, eds. The skull. Patterns of structural and systematic diversity, Vol. 3. Chicago: University of Chicago Press, 150–196. Smith ND, Makovicky PJ, Hammer WR, Currie PJ. 2007. Osteology of Cryolophosaurus ellioti (Dinosauria: Theropoda) from the Early Jurassic of Antarctica and impli- cations for early theropod evolution. Zoological Journal of the Linnean Society 151:377–421 DOI 10.1111/j.1096-3642.2007.00325.x. Stromer E. 1934. Die Zähne des Compsognathus und Bemerkungen über das Gebiß der Theropoda. Centralblatt für Mineralogie, Geologie und Paläontologie, B 1934:74–85. Sues H-D, Nesbitt SJ, Berman DS, Henrici AC. 2011. A late-surviving basal theropod dinosaur from the latest Triassic of North America. Proceedings of the Royal Society B 278:3459–3464 DOI 10.1098/rspb.2011.0410. Tallman M, Almécija S, Reber SL, Alba DM, Moyà-Solà S. 2013. The distal tibia of Hispanopithecus laietanus: more evidence for mosaic evolution in Miocene apes. Journal of Human Evolution 64:319–327 DOI 10.1016/j.jhevol.2012.07.009.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 39/41 Therrien F, Henderson DM. 2007. My theropod is bigger than yours...or not: estimat- ing body size from skull length in theropods. Journal of Vertebrate Paleontology 27:108–115 DOI 10.1671/0272-4634(2007)27[108:MTIBTY]2.0.CO;2. Tortosa T, Buffetaut E, Vialle N, Dutour Y, Turini E, Cheylan G. 2013. A new abelisaurid dinosaur from the Late Cretaceous of southern France: palaeobiogeo- graphical implications. Annales de Paléontologie 100:63–86. Tsuihiji T, Watabe M, Tsogtbaatar K, Tsubamoto T, Barsbold R, Suzuki S, Lee AH, Ridgely RC, Kawahara Y, Witmer LM. 2011. Cranial osteology of a juvenile specimens of Tarbosaurus bataar (Theropoda, Tyrannosauridae) from the Nemegt Formation (Upper Cretaceous) of Bugin Tsav, Mongolia. Journal of Vertebrate Paleontology 31:497–517 DOI 10.1080/02724634.2011.557116. Turner AH, Makovicky PJ, Norell MA. 2012. A review of dromaeosaurid systematics and paravian phylogeny. Bulletin of the American Museum of Natural History 371:1–206 DOI 10.1206/748.1. Turner AH, Pol D, Clarke JA, Erickson GM, Norell MA. 2007. A basal dromaeosaurid and size evolution preceding avian flight. Science 317:1378–1381 DOI 10.1126/science.1144066. Tykoski RS. 1998. The osteology of Syntarsus kayentakatae and its implications for ceratosaurid phylogeny. Austin: The University of Texas. Tykoski RS, Rowe TB. 2004. Ceratosauria. In: Weishampel DB, Dodson P, Osmólska H, eds. The Dinosauria. Berkeley: University of California Press, 47–70. Webster AJ, Purvis A. 2002. Testing the accuracy of methods for reconstructing ancestral states of continuous characters. Proceedings of the Royal Society of London B 269:143–149 DOI 10.1098/rspb.2001.1873. Werner J, Griebeler EM. 2014. Allometries of maximum growth rate versus body mass at maximum growth indicate that non-avian dinosaurs had growth rates typical of fast growing ectothermic sauropsids. PLoS ONE 9:e88834 DOI 10.1371/journal.pone.0088834. Wilson JA, Sereno PC. 1998. Early evolution and higher-level phylogeny of sauropod dinosaurs. Society of Vertebrate Paleontology Memoir 5:1–68 DOI 10.2307/3889325. Witmer LM. 1997. The evolution of the antorbital cavity of archosaurs: a study in soft- tissue reconstruction in the fossil record with an analysis of the function of pneu- maticity. Society of Vertebrate Paleontology Memoir 3:1–73 DOI 10.2307/3889342. Witzel U, Mannhardt J, Goessling R, Micheeli P, Preuschoft H. 2011. Finite element analyses and virtual syntheses of biological structures and their application to sauropod skulls. In: Klein N, Remes K, Gee CT, Sander PM, eds. Biology of the sauropod dinosaurs: understanding the life of giants. Bloomington: Indiana University Press, 171–181. Witzel U, Preuschoft H. 2005. Finite-element model construction for the virtual synthesis of the skulls in vertebrates: case study of Diplodocus. The Anatomical Record 283A:391–401 DOI 10.1002/ar.a.20174.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 40/41 Xu X, Clark JM, Forster CA, Norell MA, Erickson GM, Eberth DA, Jia C, Zhao Q. 2006. A basal tyrannosauroid dinosaur from the Late Jurassic of China. Nature 439:715–718 DOI 10.1038/nature04511. Xu X, Norell MA, Kuang X, Wang X, Zhao Q, Jia C. 2004. Basal tyrannosauroids from China and evidence for protofeathers in tyrannosauroids. Nature 431:680–684 DOI 10.1038/nature02855. Xu X, Wu X. 2001. Cranial morphology of Sinornithosaurus millenii Xu et al. 1999 (Dinosauria: Theropoda: Dromaeosauridae) from the Yixian Formation of Liaoning, China. Canadian Journal of Earth Sciences 38:1739–1752 DOI 10.1139/e01-082. Young MT, Larvan MD. 2010. Macroevolutionary trends in the skull of sauropodomorph dinosaurs—the largest terrestrial animals to have ever lived. In: Elewa AMT, ed. Morphometrics for non-morphometricans. Berlin: Springer, 259–269. Zanno LE, Makovicky PJ. 2011. Herbivorous ecomorphology and specialization patterns in theropod dinosaur evolution. Proceedings of the National Academy of Sciences of the United States of America 108:232–237 DOI 10.1073/pnas.1011924108. Zelditch ML, Swiderski DL, Sheets HD. 2012. Geometric morphometrics for biologists: a primer. Amsterdam: Elsevier Academic Press.

Foth et al. (2016), PeerJ, DOI 10.7717/peerj.1589 41/41